rocksdb/db/merge_helper.cc

793 lines
31 KiB
C++
Raw Normal View History

// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
// This source code is licensed under both the GPLv2 (found in the
// COPYING file in the root directory) and Apache 2.0 License
// (found in the LICENSE.Apache file in the root directory).
#include "db/merge_helper.h"
#include <string>
#include "db/blob/blob_fetcher.h"
#include "db/blob/blob_index.h"
#include "db/blob/prefetch_buffer_collection.h"
#include "db/compaction/compaction_iteration_stats.h"
#include "db/dbformat.h"
#include "db/wide/wide_column_serialization.h"
#include "db/wide/wide_columns_helper.h"
#include "logging/logging.h"
#include "monitoring/perf_context_imp.h"
#include "monitoring/statistics_impl.h"
#include "port/likely.h"
#include "rocksdb/comparator.h"
#include "rocksdb/db.h"
#include "rocksdb/merge_operator.h"
#include "rocksdb/system_clock.h"
#include "table/format.h"
#include "table/internal_iterator.h"
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
#include "util/overload.h"
namespace ROCKSDB_NAMESPACE {
MergeHelper::MergeHelper(Env* env, const Comparator* user_comparator,
const MergeOperator* user_merge_operator,
const CompactionFilter* compaction_filter,
Logger* logger, bool assert_valid_internal_key,
SequenceNumber latest_snapshot,
const SnapshotChecker* snapshot_checker, int level,
Statistics* stats,
const std::atomic<bool>* shutting_down)
: env_(env),
clock_(env->GetSystemClock().get()),
user_comparator_(user_comparator),
user_merge_operator_(user_merge_operator),
compaction_filter_(compaction_filter),
shutting_down_(shutting_down),
logger_(logger),
assert_valid_internal_key_(assert_valid_internal_key),
allow_single_operand_(false),
latest_snapshot_(latest_snapshot),
snapshot_checker_(snapshot_checker),
level_(level),
keys_(),
filter_timer_(clock_),
total_filter_time_(0U),
stats_(stats) {
assert(user_comparator_ != nullptr);
if (user_merge_operator_) {
allow_single_operand_ = user_merge_operator_->AllowSingleOperand();
}
}
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
template <typename Visitor>
Status MergeHelper::TimedFullMergeCommonImpl(
const MergeOperator* merge_operator, const Slice& key,
MergeOperator::MergeOperationInputV3::ExistingValue&& existing_value,
const std::vector<Slice>& operands, Logger* logger, Statistics* statistics,
SystemClock* clock, bool update_num_ops_stats,
MergeOperator::OpFailureScope* op_failure_scope, Visitor&& visitor) {
assert(merge_operator);
assert(!operands.empty());
if (update_num_ops_stats) {
RecordInHistogram(statistics, READ_NUM_MERGE_OPERANDS,
static_cast<uint64_t>(operands.size()));
}
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
const MergeOperator::MergeOperationInputV3 merge_in(
key, std::move(existing_value), operands, logger);
MergeOperator::MergeOperationOutputV3 merge_out;
bool success = false;
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
{
StopWatchNano timer(clock, statistics != nullptr);
PERF_TIMER_GUARD(merge_operator_time_nanos);
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
success = merge_operator->FullMergeV3(merge_in, &merge_out);
Remove wasteful instrumentation in FullMerge (stacked on D59577) Summary: [ This diff is stacked on top of D59577 ] We keep calling timer.ElapsedNanos() on every call to MergeOperator::FullMerge even when statistics are disabled, this is wasteful. I run the readseq benchmark on a DB containing 100K merge operands for 100K keys (1 operand per key) with 1GB block cache I see slight performance improvment Original results ``` $ ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=100000 --num=100000 --db="/dev/shm/100K_merge_compacted/" --cache_size=1073741824 --use_existing_db --disable_auto_compactions ------------------------------------------------ DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.498 micros/op 2006597 ops/sec; 222.0 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.295 micros/op 3393627 ops/sec; 375.4 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.285 micros/op 3511155 ops/sec; 388.4 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.286 micros/op 3500470 ops/sec; 387.2 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.283 micros/op 3530751 ops/sec; 390.6 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.289 micros/op 3464811 ops/sec; 383.3 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.277 micros/op 3612814 ops/sec; 399.7 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.283 micros/op 3539640 ops/sec; 391.6 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.285 micros/op 3503766 ops/sec; 387.6 MB/s ``` After patch ``` $ ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=100000 --num=100000 --db="/dev/shm/100K_merge_compacted/" --cache_size=1073741824 --use_existing_db --disable_auto_compactions ------------------------------------------------ DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.476 micros/op 2100119 ops/sec; 232.3 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.278 micros/op 3600887 ops/sec; 398.4 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.275 micros/op 3636698 ops/sec; 402.3 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.271 micros/op 3691661 ops/sec; 408.4 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.273 micros/op 3661534 ops/sec; 405.1 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.276 micros/op 3627106 ops/sec; 401.3 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.272 micros/op 3682635 ops/sec; 407.4 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.266 micros/op 3758331 ops/sec; 415.8 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.266 micros/op 3761907 ops/sec; 416.2 MB/s ``` Test Plan: make check -j64 Reviewers: yhchiang, sdong Reviewed By: sdong Subscribers: andrewkr, dhruba Differential Revision: https://reviews.facebook.net/D59583
2016-06-13 23:22:14 +00:00
RecordTick(statistics, MERGE_OPERATION_TOTAL_TIME,
statistics ? timer.ElapsedNanos() : 0);
}
if (!success) {
RecordTick(statistics, NUMBER_MERGE_FAILURES);
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
if (op_failure_scope) {
*op_failure_scope = merge_out.op_failure_scope;
// Apply default per merge_operator.h
if (*op_failure_scope == MergeOperator::OpFailureScope::kDefault) {
*op_failure_scope = MergeOperator::OpFailureScope::kTryMerge;
}
}
return Status::Corruption(Status::SubCode::kMergeOperatorFailed);
}
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
return std::visit(std::forward<Visitor>(visitor),
std::move(merge_out.new_value));
}
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
Status MergeHelper::TimedFullMergeImpl(
const MergeOperator* merge_operator, const Slice& key,
MergeOperator::MergeOperationInputV3::ExistingValue&& existing_value,
const std::vector<Slice>& operands, Logger* logger, Statistics* statistics,
SystemClock* clock, bool update_num_ops_stats, std::string* result,
Slice* result_operand, ValueType* result_type,
MergeOperator::OpFailureScope* op_failure_scope) {
assert(result);
assert(result_type);
auto visitor = overload{
[&](std::string&& new_value) -> Status {
*result_type = kTypeValue;
if (result_operand) {
*result_operand = Slice(nullptr, 0);
}
*result = std::move(new_value);
return Status::OK();
},
[&](MergeOperator::MergeOperationOutputV3::NewColumns&& new_columns)
-> Status {
*result_type = kTypeWideColumnEntity;
if (result_operand) {
*result_operand = Slice(nullptr, 0);
}
result->clear();
WideColumns sorted_columns;
sorted_columns.reserve(new_columns.size());
for (const auto& column : new_columns) {
sorted_columns.emplace_back(column.first, column.second);
}
WideColumnsHelper::SortColumns(sorted_columns);
return WideColumnSerialization::Serialize(sorted_columns, *result);
},
[&](Slice&& operand) -> Status {
*result_type = kTypeValue;
if (result_operand) {
*result_operand = operand;
result->clear();
} else {
result->assign(operand.data(), operand.size());
}
return Status::OK();
}};
return TimedFullMergeCommonImpl(merge_operator, key,
std::move(existing_value), operands, logger,
statistics, clock, update_num_ops_stats,
op_failure_scope, std::move(visitor));
}
Status MergeHelper::TimedFullMergeImpl(
const MergeOperator* merge_operator, const Slice& key,
MergeOperator::MergeOperationInputV3::ExistingValue&& existing_value,
const std::vector<Slice>& operands, Logger* logger, Statistics* statistics,
SystemClock* clock, bool update_num_ops_stats, std::string* result_value,
PinnableWideColumns* result_entity,
MergeOperator::OpFailureScope* op_failure_scope) {
assert(result_value || result_entity);
assert(!result_value || !result_entity);
auto visitor = overload{
[&](std::string&& new_value) -> Status {
if (result_value) {
*result_value = std::move(new_value);
return Status::OK();
}
assert(result_entity);
result_entity->SetPlainValue(std::move(new_value));
return Status::OK();
},
[&](MergeOperator::MergeOperationOutputV3::NewColumns&& new_columns)
-> Status {
if (result_value) {
if (!new_columns.empty() &&
new_columns.front().first == kDefaultWideColumnName) {
*result_value = std::move(new_columns.front().second);
} else {
result_value->clear();
}
return Status::OK();
}
assert(result_entity);
WideColumns sorted_columns;
sorted_columns.reserve(new_columns.size());
for (const auto& column : new_columns) {
sorted_columns.emplace_back(column.first, column.second);
}
WideColumnsHelper::SortColumns(sorted_columns);
std::string result;
const Status s =
WideColumnSerialization::Serialize(sorted_columns, result);
if (!s.ok()) {
result_entity->Reset();
return s;
}
return result_entity->SetWideColumnValue(std::move(result));
},
[&](Slice&& operand) -> Status {
if (result_value) {
result_value->assign(operand.data(), operand.size());
return Status::OK();
}
assert(result_entity);
result_entity->SetPlainValue(operand);
return Status::OK();
}};
return TimedFullMergeCommonImpl(merge_operator, key,
std::move(existing_value), operands, logger,
statistics, clock, update_num_ops_stats,
op_failure_scope, std::move(visitor));
}
Status MergeHelper::TimedFullMerge(
const MergeOperator* merge_operator, const Slice& key, NoBaseValueTag,
const std::vector<Slice>& operands, Logger* logger, Statistics* statistics,
SystemClock* clock, bool update_num_ops_stats, std::string* result,
Slice* result_operand, ValueType* result_type,
MergeOperator::OpFailureScope* op_failure_scope) {
MergeOperator::MergeOperationInputV3::ExistingValue existing_value;
return TimedFullMergeImpl(merge_operator, key, std::move(existing_value),
operands, logger, statistics, clock,
update_num_ops_stats, result, result_operand,
result_type, op_failure_scope);
}
Status MergeHelper::TimedFullMerge(
const MergeOperator* merge_operator, const Slice& key, PlainBaseValueTag,
const Slice& value, const std::vector<Slice>& operands, Logger* logger,
Statistics* statistics, SystemClock* clock, bool update_num_ops_stats,
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
std::string* result, Slice* result_operand, ValueType* result_type,
MergeOperator::OpFailureScope* op_failure_scope) {
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
MergeOperator::MergeOperationInputV3::ExistingValue existing_value(value);
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
return TimedFullMergeImpl(merge_operator, key, std::move(existing_value),
operands, logger, statistics, clock,
update_num_ops_stats, result, result_operand,
result_type, op_failure_scope);
}
Status MergeHelper::TimedFullMerge(
const MergeOperator* merge_operator, const Slice& key, WideBaseValueTag,
const Slice& entity, const std::vector<Slice>& operands, Logger* logger,
Statistics* statistics, SystemClock* clock, bool update_num_ops_stats,
std::string* result, Slice* result_operand, ValueType* result_type,
MergeOperator::OpFailureScope* op_failure_scope) {
MergeOperator::MergeOperationInputV3::ExistingValue existing_value;
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
Slice entity_copy(entity);
WideColumns existing_columns;
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
const Status s =
WideColumnSerialization::Deserialize(entity_copy, existing_columns);
if (!s.ok()) {
return s;
}
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
existing_value = std::move(existing_columns);
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
return TimedFullMergeImpl(merge_operator, key, std::move(existing_value),
operands, logger, statistics, clock,
update_num_ops_stats, result, result_operand,
result_type, op_failure_scope);
}
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
Status MergeHelper::TimedFullMerge(
const MergeOperator* merge_operator, const Slice& key, NoBaseValueTag,
const std::vector<Slice>& operands, Logger* logger, Statistics* statistics,
SystemClock* clock, bool update_num_ops_stats, std::string* result_value,
PinnableWideColumns* result_entity,
MergeOperator::OpFailureScope* op_failure_scope) {
MergeOperator::MergeOperationInputV3::ExistingValue existing_value;
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
return TimedFullMergeImpl(merge_operator, key, std::move(existing_value),
operands, logger, statistics, clock,
update_num_ops_stats, result_value, result_entity,
op_failure_scope);
}
Status MergeHelper::TimedFullMerge(
const MergeOperator* merge_operator, const Slice& key, PlainBaseValueTag,
const Slice& value, const std::vector<Slice>& operands, Logger* logger,
Statistics* statistics, SystemClock* clock, bool update_num_ops_stats,
std::string* result_value, PinnableWideColumns* result_entity,
MergeOperator::OpFailureScope* op_failure_scope) {
MergeOperator::MergeOperationInputV3::ExistingValue existing_value(value);
return TimedFullMergeImpl(merge_operator, key, std::move(existing_value),
operands, logger, statistics, clock,
update_num_ops_stats, result_value, result_entity,
op_failure_scope);
}
Status MergeHelper::TimedFullMerge(
const MergeOperator* merge_operator, const Slice& key, WideBaseValueTag,
const Slice& entity, const std::vector<Slice>& operands, Logger* logger,
Statistics* statistics, SystemClock* clock, bool update_num_ops_stats,
std::string* result_value, PinnableWideColumns* result_entity,
MergeOperator::OpFailureScope* op_failure_scope) {
MergeOperator::MergeOperationInputV3::ExistingValue existing_value;
Slice entity_copy(entity);
WideColumns existing_columns;
const Status s =
WideColumnSerialization::Deserialize(entity_copy, existing_columns);
if (!s.ok()) {
return s;
}
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
existing_value = std::move(existing_columns);
return TimedFullMergeImpl(merge_operator, key, std::move(existing_value),
operands, logger, statistics, clock,
update_num_ops_stats, result_value, result_entity,
op_failure_scope);
}
// PRE: iter points to the first merge type entry
// POST: iter points to the first entry beyond the merge process (or the end)
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
// keys_, operands_ are updated to reflect the merge result.
// keys_ stores the list of keys encountered while merging.
// operands_ stores the list of merge operands encountered while merging.
// keys_[i] corresponds to operands_[i] for each i.
//
// TODO: Avoid the snapshot stripe map lookup in CompactionRangeDelAggregator
// and just pass the StripeRep corresponding to the stripe being merged.
Status MergeHelper::MergeUntil(InternalIterator* iter,
CompactionRangeDelAggregator* range_del_agg,
const SequenceNumber stop_before,
const bool at_bottom,
const bool allow_data_in_errors,
const BlobFetcher* blob_fetcher,
const std::string* const full_history_ts_low,
PrefetchBufferCollection* prefetch_buffers,
CompactionIterationStats* c_iter_stats) {
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
// Get a copy of the internal key, before it's invalidated by iter->Next()
// Also maintain the list of merge operands seen.
assert(HasOperator());
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
keys_.clear();
Introduce FullMergeV2 (eliminate memcpy from merge operators) Summary: This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice> This diff is stacked on top of D56493 and D56511 In this diff we - Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future - Replace std::deque<std::string> with std::vector<Slice> to pass operands - Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187) - Allow FullMergeV2 output to be an existing operand ``` [Everything in Memtable | 10K operands | 10 KB each | 1 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s [master] readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s ``` ``` [Everything in Memtable | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s [master] readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 1 operand per key] [FullMergeV2] $ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s [master] readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions [FullMergeV2] readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s [master] readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s ``` Test Plan: COMPILE_WITH_ASAN=1 make check -j64 Reviewers: yhchiang, andrewkr, sdong Reviewed By: sdong Subscribers: lovro, andrewkr, dhruba Differential Revision: https://reviews.facebook.net/D57075
2016-07-20 16:49:03 +00:00
merge_context_.Clear();
has_compaction_filter_skip_until_ = false;
assert(user_merge_operator_);
assert(user_comparator_);
const size_t ts_sz = user_comparator_->timestamp_size();
if (full_history_ts_low) {
assert(ts_sz > 0);
assert(ts_sz == full_history_ts_low->size());
}
bool first_key = true;
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
// We need to parse the internal key again as the parsed key is
// backed by the internal key!
// Assume no internal key corruption as it has been successfully parsed
// by the caller.
// original_key_is_iter variable is just caching the information:
// original_key_is_iter == (iter->key().ToString() == original_key)
bool original_key_is_iter = true;
std::string original_key = iter->key().ToString();
// Important:
// orig_ikey is backed by original_key if keys_.empty()
// orig_ikey is backed by keys_.back() if !keys_.empty()
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
ParsedInternalKey orig_ikey;
Status s = ParseInternalKey(original_key, &orig_ikey, allow_data_in_errors);
assert(s.ok());
if (!s.ok()) return s;
assert(kTypeMerge == orig_ikey.type);
bool hit_the_next_user_key = false;
int cmp_with_full_history_ts_low = 0;
for (; iter->Valid(); iter->Next(), original_key_is_iter = false) {
if (IsShuttingDown()) {
s = Status::ShutdownInProgress();
return s;
}
Refactor AddRangeDels() + consider range tombstone during compaction file cutting (#11113) Summary: A second attempt after https://github.com/facebook/rocksdb/issues/10802, with bug fixes and refactoring. This PR updates compaction logic to take range tombstones into account when determining whether to cut the current compaction output file (https://github.com/facebook/rocksdb/issues/4811). Before this change, only point keys were considered, and range tombstones could cause large compactions. For example, if the current compaction outputs is a range tombstone [a, b) and 2 point keys y, z, they would be added to the same file, and may overlap with too many files in the next level and cause a large compaction in the future. This PR also includes ajkr's effort to simplify the logic to add range tombstones to compaction output files in `AddRangeDels()` ([https://github.com/facebook/rocksdb/issues/11078](https://github.com/facebook/rocksdb/pull/11078#issuecomment-1386078861)). The main change is for `CompactionIterator` to emit range tombstone start keys to be processed by `CompactionOutputs`. A new class `CompactionMergingIterator` is introduced to replace `MergingIterator` under `CompactionIterator` to enable emitting of range tombstone start keys. Further improvement after this PR include cutting compaction output at some grandparent boundary key (instead of the next output key) when cutting within a range tombstone to reduce overlap with grandparents. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11113 Test Plan: * added unit test in db_range_del_test * crash test with a small key range: `python3 tools/db_crashtest.py blackbox --simple --max_key=100 --interval=600 --write_buffer_size=262144 --target_file_size_base=256 --max_bytes_for_level_base=262144 --block_size=128 --value_size_mult=33 --subcompactions=10 --use_multiget=1 --delpercent=3 --delrangepercent=2 --verify_iterator_with_expected_state_one_in=2 --num_iterations=10` Reviewed By: ajkr Differential Revision: D42655709 Pulled By: cbi42 fbshipit-source-id: 8367e36ef5640e8f21c14a3855d4a8d6e360a34c
2023-02-22 20:28:18 +00:00
// Skip range tombstones emitted by the compaction iterator.
if (iter->IsDeleteRangeSentinelKey()) {
continue;
}
ParsedInternalKey ikey;
Introduce FullMergeV2 (eliminate memcpy from merge operators) Summary: This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice> This diff is stacked on top of D56493 and D56511 In this diff we - Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future - Replace std::deque<std::string> with std::vector<Slice> to pass operands - Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187) - Allow FullMergeV2 output to be an existing operand ``` [Everything in Memtable | 10K operands | 10 KB each | 1 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s [master] readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s ``` ``` [Everything in Memtable | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s [master] readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 1 operand per key] [FullMergeV2] $ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s [master] readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions [FullMergeV2] readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s [master] readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s ``` Test Plan: COMPILE_WITH_ASAN=1 make check -j64 Reviewers: yhchiang, andrewkr, sdong Reviewed By: sdong Subscribers: lovro, andrewkr, dhruba Differential Revision: https://reviews.facebook.net/D57075
2016-07-20 16:49:03 +00:00
assert(keys_.size() == merge_context_.GetNumOperands());
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
Status pik_status =
ParseInternalKey(iter->key(), &ikey, allow_data_in_errors);
Slice ts;
if (pik_status.ok()) {
ts = ExtractTimestampFromUserKey(ikey.user_key, ts_sz);
if (full_history_ts_low) {
cmp_with_full_history_ts_low =
user_comparator_->CompareTimestamp(ts, *full_history_ts_low);
}
}
if (!pik_status.ok()) {
// stop at corrupted key
if (assert_valid_internal_key_) {
return pik_status;
}
break;
} else if (first_key) {
// If user-defined timestamp is enabled, we expect both user key and
// timestamps are equal, as a sanity check.
assert(user_comparator_->Equal(ikey.user_key, orig_ikey.user_key));
first_key = false;
} else if (!user_comparator_->EqualWithoutTimestamp(ikey.user_key,
orig_ikey.user_key) ||
(ts_sz > 0 &&
!user_comparator_->Equal(ikey.user_key, orig_ikey.user_key) &&
cmp_with_full_history_ts_low >= 0)) {
// 1) hit a different user key, or
// 2) user-defined timestamp is enabled, and hit a version of user key NOT
// eligible for GC, then stop right here.
hit_the_next_user_key = true;
break;
} else if (stop_before > 0 && ikey.sequence <= stop_before &&
LIKELY(snapshot_checker_ == nullptr ||
snapshot_checker_->CheckInSnapshot(ikey.sequence,
stop_before) !=
SnapshotCheckerResult::kNotInSnapshot)) {
// hit an entry that's possibly visible by the previous snapshot, can't
// touch that
break;
}
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
// At this point we are guaranteed that we need to process this key.
Support for SingleDelete() Summary: This patch fixes #7460559. It introduces SingleDelete as a new database operation. This operation can be used to delete keys that were never overwritten (no put following another put of the same key). If an overwritten key is single deleted the behavior is undefined. Single deletion of a non-existent key has no effect but multiple consecutive single deletions are not allowed (see limitations). In contrast to the conventional Delete() operation, the deletion entry is removed along with the value when the two are lined up in a compaction. Note: The semantics are similar to @igor's prototype that allowed to have this behavior on the granularity of a column family ( https://reviews.facebook.net/D42093 ). This new patch, however, is more aggressive when it comes to removing tombstones: It removes the SingleDelete together with the value whenever there is no snapshot between them while the older patch only did this when the sequence number of the deletion was older than the earliest snapshot. Most of the complex additions are in the Compaction Iterator, all other changes should be relatively straightforward. The patch also includes basic support for single deletions in db_stress and db_bench. Limitations: - Not compatible with cuckoo hash tables - Single deletions cannot be used in combination with merges and normal deletions on the same key (other keys are not affected by this) - Consecutive single deletions are currently not allowed (and older version of this patch supported this so it could be resurrected if needed) Test Plan: make all check Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor Reviewed By: igor Subscribers: maykov, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 18:42:56 +00:00
assert(IsValueType(ikey.type));
if (ikey.type != kTypeMerge) {
// hit a put/delete/single delete
// => merge the put value or a nullptr with operands_
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
// => store result in operands_.back() (and update keys_.back())
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
// => change the entry type for keys_.back()
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
// We are done! Success!
// If there are no operands, just return the Status::OK(). That will cause
// the compaction iterator to write out the key we're currently at, which
// is the put/delete we just encountered.
if (keys_.empty()) {
return s;
}
// TODO: if we're in compaction and it's a put, it would be nice to run
// compaction filter on it.
std::string merge_result;
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
ValueType merge_result_type;
MergeOperator::OpFailureScope op_failure_scope;
if (range_del_agg &&
range_del_agg->ShouldDelete(
ikey, RangeDelPositioningMode::kForwardTraversal)) {
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
s = TimedFullMerge(user_merge_operator_, ikey.user_key, kNoBaseValue,
merge_context_.GetOperands(), logger_, stats_,
clock_, /* update_num_ops_stats */ false,
&merge_result, /* result_operand */ nullptr,
&merge_result_type, &op_failure_scope);
} else if (ikey.type == kTypeValue) {
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
s = TimedFullMerge(user_merge_operator_, ikey.user_key, kPlainBaseValue,
iter->value(), merge_context_.GetOperands(), logger_,
stats_, clock_, /* update_num_ops_stats */ false,
&merge_result, /* result_operand */ nullptr,
&merge_result_type, &op_failure_scope);
} else if (ikey.type == kTypeBlobIndex) {
BlobIndex blob_index;
s = blob_index.DecodeFrom(iter->value());
if (!s.ok()) {
return s;
}
FilePrefetchBuffer* prefetch_buffer =
prefetch_buffers ? prefetch_buffers->GetOrCreatePrefetchBuffer(
blob_index.file_number())
: nullptr;
uint64_t bytes_read = 0;
assert(blob_fetcher);
PinnableSlice blob_value;
s = blob_fetcher->FetchBlob(ikey.user_key, blob_index, prefetch_buffer,
&blob_value, &bytes_read);
if (!s.ok()) {
return s;
}
if (c_iter_stats) {
++c_iter_stats->num_blobs_read;
c_iter_stats->total_blob_bytes_read += bytes_read;
}
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
s = TimedFullMerge(user_merge_operator_, ikey.user_key, kPlainBaseValue,
blob_value, merge_context_.GetOperands(), logger_,
stats_, clock_, /* update_num_ops_stats */ false,
&merge_result, /* result_operand */ nullptr,
&merge_result_type, &op_failure_scope);
} else if (ikey.type == kTypeWideColumnEntity) {
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
s = TimedFullMerge(user_merge_operator_, ikey.user_key, kWideBaseValue,
iter->value(), merge_context_.GetOperands(), logger_,
stats_, clock_, /* update_num_ops_stats */ false,
&merge_result, /* result_operand */ nullptr,
&merge_result_type, &op_failure_scope);
} else {
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
s = TimedFullMerge(user_merge_operator_, ikey.user_key, kNoBaseValue,
merge_context_.GetOperands(), logger_, stats_,
clock_, /* update_num_ops_stats */ false,
&merge_result, /* result_operand */ nullptr,
&merge_result_type, &op_failure_scope);
}
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
// We store the result in keys_.back() and operands_.back()
// if nothing went wrong (i.e.: no operand corruption on disk)
if (s.ok()) {
// The original key encountered
original_key = std::move(keys_.back());
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
assert(merge_result_type == kTypeValue ||
merge_result_type == kTypeWideColumnEntity);
orig_ikey.type = merge_result_type;
UpdateInternalKey(&original_key, orig_ikey.sequence, orig_ikey.type);
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
keys_.clear();
Introduce FullMergeV2 (eliminate memcpy from merge operators) Summary: This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice> This diff is stacked on top of D56493 and D56511 In this diff we - Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future - Replace std::deque<std::string> with std::vector<Slice> to pass operands - Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187) - Allow FullMergeV2 output to be an existing operand ``` [Everything in Memtable | 10K operands | 10 KB each | 1 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s [master] readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s ``` ``` [Everything in Memtable | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s [master] readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 1 operand per key] [FullMergeV2] $ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s [master] readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions [FullMergeV2] readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s [master] readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s ``` Test Plan: COMPILE_WITH_ASAN=1 make check -j64 Reviewers: yhchiang, andrewkr, sdong Reviewed By: sdong Subscribers: lovro, andrewkr, dhruba Differential Revision: https://reviews.facebook.net/D57075
2016-07-20 16:49:03 +00:00
merge_context_.Clear();
keys_.emplace_front(std::move(original_key));
Introduce FullMergeV2 (eliminate memcpy from merge operators) Summary: This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice> This diff is stacked on top of D56493 and D56511 In this diff we - Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future - Replace std::deque<std::string> with std::vector<Slice> to pass operands - Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187) - Allow FullMergeV2 output to be an existing operand ``` [Everything in Memtable | 10K operands | 10 KB each | 1 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s [master] readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s ``` ``` [Everything in Memtable | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s [master] readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 1 operand per key] [FullMergeV2] $ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s [master] readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions [FullMergeV2] readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s [master] readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s ``` Test Plan: COMPILE_WITH_ASAN=1 make check -j64 Reviewers: yhchiang, andrewkr, sdong Reviewed By: sdong Subscribers: lovro, andrewkr, dhruba Differential Revision: https://reviews.facebook.net/D57075
2016-07-20 16:49:03 +00:00
merge_context_.PushOperand(merge_result);
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
// move iter to the next entry
iter->Next();
} else if (op_failure_scope ==
MergeOperator::OpFailureScope::kMustMerge) {
// Change to `Status::MergeInProgress()` to denote output consists of
// merge operands only. Leave `iter` at the non-merge entry so it will
// be output after.
s = Status::MergeInProgress();
}
return s;
} else {
// hit a merge
// => if there is a compaction filter, apply it.
Compaction Support for Range Deletion Summary: This diff introduces RangeDelAggregator, which takes ownership of iterators provided to it via AddTombstones(). The tombstones are organized in a two-level map (snapshot stripe -> begin key -> tombstone). Tombstone creation avoids data copy by holding Slices returned by the iterator, which remain valid thanks to pinning. For compaction, we create a hierarchical range tombstone iterator with structure matching the iterator over compaction input data. An aggregator based on that iterator is used by CompactionIterator to determine which keys are covered by range tombstones. In case of merge operand, the same aggregator is used by MergeHelper. Upon finishing each file in the compaction, relevant range tombstones are added to the output file's range tombstone metablock and file boundaries are updated accordingly. To check whether a key is covered by range tombstone, RangeDelAggregator::ShouldDelete() considers tombstones in the key's snapshot stripe. When this function is used outside of compaction, it also checks newer stripes, which can contain covering tombstones. Currently the intra-stripe check involves a linear scan; however, in the future we plan to collapse ranges within a stripe such that binary search can be used. RangeDelAggregator::AddToBuilder() adds all range tombstones in the table's key-range to a new table's range tombstone meta-block. Since range tombstones may fall in the gap between files, we may need to extend some files' key-ranges. The strategy is (1) first file extends as far left as possible and other files do not extend left, (2) all files extend right until either the start of the next file or the end of the last range tombstone in the gap, whichever comes first. One other notable change is adding release/move semantics to ScopedArenaIterator such that it can be used to transfer ownership of an arena-allocated iterator, similar to how unique_ptr is used for malloc'd data. Depends on D61473 Test Plan: compaction_iterator_test, mock_table, end-to-end tests in D63927 Reviewers: sdong, IslamAbdelRahman, wanning, yhchiang, lightmark Reviewed By: lightmark Subscribers: andrewkr, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D62205
2016-10-18 19:04:56 +00:00
// => check for range tombstones covering the operand
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
// => merge the operand into the front of the operands_ list
// if not filtered
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
// => then continue because we haven't yet seen a Put/Delete.
//
// Keep queuing keys and operands until we either meet a put / delete
// request or later did a partial merge.
Slice value_slice = iter->value();
// add an operand to the list if:
// 1) it's included in one of the snapshots. in that case we *must* write
// it out, no matter what compaction filter says
// 2) it's not filtered by a compaction filter
CompactionFilter::Decision filter =
ikey.sequence <= latest_snapshot_
? CompactionFilter::Decision::kKeep
: FilterMerge(orig_ikey.user_key, value_slice);
if (filter != CompactionFilter::Decision::kRemoveAndSkipUntil &&
range_del_agg != nullptr &&
range_del_agg->ShouldDelete(
iter->key(), RangeDelPositioningMode::kForwardTraversal)) {
filter = CompactionFilter::Decision::kRemove;
}
if (filter == CompactionFilter::Decision::kKeep ||
filter == CompactionFilter::Decision::kChangeValue) {
if (original_key_is_iter) {
// this is just an optimization that saves us one memcpy
keys_.emplace_front(original_key);
} else {
keys_.emplace_front(iter->key().ToString());
}
if (keys_.size() == 1) {
// we need to re-anchor the orig_ikey because it was anchored by
// original_key before
pik_status =
ParseInternalKey(keys_.back(), &orig_ikey, allow_data_in_errors);
pik_status.PermitUncheckedError();
assert(pik_status.ok());
}
if (filter == CompactionFilter::Decision::kKeep) {
merge_context_.PushOperand(
value_slice, iter->IsValuePinned() /* operand_pinned */);
} else {
assert(filter == CompactionFilter::Decision::kChangeValue);
// Compaction filter asked us to change the operand from value_slice
// to compaction_filter_value_.
merge_context_.PushOperand(compaction_filter_value_, false);
}
} else if (filter == CompactionFilter::Decision::kRemoveAndSkipUntil) {
// Compaction filter asked us to remove this key altogether
// (not just this operand), along with some keys following it.
keys_.clear();
merge_context_.Clear();
has_compaction_filter_skip_until_ = true;
return s;
}
}
}
if (cmp_with_full_history_ts_low >= 0) {
size_t num_merge_operands = merge_context_.GetNumOperands();
if (ts_sz && num_merge_operands > 1) {
// We do not merge merge operands with different timestamps if they are
// not eligible for GC.
ROCKS_LOG_ERROR(logger_, "ts_sz=%d, %d merge oprands",
static_cast<int>(ts_sz),
static_cast<int>(num_merge_operands));
assert(false);
}
}
Introduce FullMergeV2 (eliminate memcpy from merge operators) Summary: This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice> This diff is stacked on top of D56493 and D56511 In this diff we - Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future - Replace std::deque<std::string> with std::vector<Slice> to pass operands - Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187) - Allow FullMergeV2 output to be an existing operand ``` [Everything in Memtable | 10K operands | 10 KB each | 1 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s [master] readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s ``` ``` [Everything in Memtable | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s [master] readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 1 operand per key] [FullMergeV2] $ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s [master] readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions [FullMergeV2] readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s [master] readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s ``` Test Plan: COMPILE_WITH_ASAN=1 make check -j64 Reviewers: yhchiang, andrewkr, sdong Reviewed By: sdong Subscribers: lovro, andrewkr, dhruba Differential Revision: https://reviews.facebook.net/D57075
2016-07-20 16:49:03 +00:00
if (merge_context_.GetNumOperands() == 0) {
// we filtered out all the merge operands
return s;
}
// We are sure we have seen this key's entire history if:
// at_bottom == true (this does not necessarily mean it is the bottommost
// layer, but rather that we are confident the key does not appear on any of
// the lower layers, at_bottom == false doesn't mean it does appear, just
// that we can't be sure, see Compaction::IsBottommostLevel for details)
// AND
// we have either encountered another key or end of key history on this
// layer.
// Note that if user-defined timestamp is enabled, we need some extra caution
// here: if full_history_ts_low is nullptr, or it's not null but the key's
// timestamp is greater than or equal to full_history_ts_low, it means this
// key cannot be dropped. We may not have seen the beginning of the key.
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
//
// When these conditions are true we are able to merge all the keys
// using full merge.
//
// For these cases we are not sure about, we simply miss the opportunity
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
// to combine the keys. Since VersionSet::SetupOtherInputs() always makes
// sure that all merge-operands on the same level get compacted together,
// this will simply lead to these merge operands moving to the next level.
bool surely_seen_the_beginning =
(hit_the_next_user_key || !iter->Valid()) && at_bottom &&
(ts_sz == 0 || cmp_with_full_history_ts_low < 0);
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
if (surely_seen_the_beginning) {
// do a final merge with nullptr as the existing value and say
// bye to the merge type (it's now converted to a Put)
assert(kTypeMerge == orig_ikey.type);
Introduce FullMergeV2 (eliminate memcpy from merge operators) Summary: This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice> This diff is stacked on top of D56493 and D56511 In this diff we - Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future - Replace std::deque<std::string> with std::vector<Slice> to pass operands - Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187) - Allow FullMergeV2 output to be an existing operand ``` [Everything in Memtable | 10K operands | 10 KB each | 1 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s [master] readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s ``` ``` [Everything in Memtable | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s [master] readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 1 operand per key] [FullMergeV2] $ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s [master] readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions [FullMergeV2] readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s [master] readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s ``` Test Plan: COMPILE_WITH_ASAN=1 make check -j64 Reviewers: yhchiang, andrewkr, sdong Reviewed By: sdong Subscribers: lovro, andrewkr, dhruba Differential Revision: https://reviews.facebook.net/D57075
2016-07-20 16:49:03 +00:00
assert(merge_context_.GetNumOperands() >= 1);
assert(merge_context_.GetNumOperands() == keys_.size());
std::string merge_result;
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
ValueType merge_result_type;
MergeOperator::OpFailureScope op_failure_scope;
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
s = TimedFullMerge(user_merge_operator_, orig_ikey.user_key, kNoBaseValue,
merge_context_.GetOperands(), logger_, stats_, clock_,
/* update_num_ops_stats */ false, &merge_result,
/* result_operand */ nullptr, &merge_result_type,
&op_failure_scope);
if (s.ok()) {
// The original key encountered
// We are certain that keys_ is not empty here (see assertions couple of
// lines before).
original_key = std::move(keys_.back());
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
assert(merge_result_type == kTypeValue ||
merge_result_type == kTypeWideColumnEntity);
orig_ikey.type = merge_result_type;
UpdateInternalKey(&original_key, orig_ikey.sequence, orig_ikey.type);
Integrate FullMergeV3 into the query and compaction paths (#11858) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11858 The patch builds on https://github.com/facebook/rocksdb/pull/11807 and integrates the `FullMergeV3` API into the read and compaction code paths by updating and extending the logic in `MergeHelper`. In particular, when it comes to merge inputs, the existing `TimedFullMergeWithEntity` is folded into `TimedFullMerge`, since wide-column base values are now handled the same way as plain base values (or no base values for that matter), e.g. they are passed directly to the `MergeOperator`. On the other hand, there is some new differentiation on the output side. Namely, there are now two sets of `TimedFullMerge` variants: one set for contexts where the complete merge result and its value type are needed (used by iterators and compactions), and another set where the merge result is needed in a form determined by the client (used by the point lookup APIs, where e.g. for `Get` we have to extract the value of the default column of any wide-column results). Implementation-wise, the two sets of overloads use different visitors to process the `std::variant` produced by `FullMergeV3`. This has the benefit of eliminating some repeated code e.g. in the point lookup paths, since `TimedFullMerge` now populates the application's result object (`PinnableSlice`/`string` or `PinnableWideColumns`) directly. Moreover, within each set of variants, there is a separate overload for the no base value/plain base value/wide-column base value cases, which eliminates some repeated branching w/r/t to the type of the base value if any. Reviewed By: jaykorean Differential Revision: D49352562 fbshipit-source-id: c2fb9853dba3fbbc6918665bde4195c4ea150a0c
2023-09-20 00:27:04 +00:00
keys_.clear();
Introduce FullMergeV2 (eliminate memcpy from merge operators) Summary: This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice> This diff is stacked on top of D56493 and D56511 In this diff we - Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future - Replace std::deque<std::string> with std::vector<Slice> to pass operands - Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187) - Allow FullMergeV2 output to be an existing operand ``` [Everything in Memtable | 10K operands | 10 KB each | 1 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s [master] readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s ``` ``` [Everything in Memtable | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s [master] readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 1 operand per key] [FullMergeV2] $ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s [master] readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions [FullMergeV2] readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s [master] readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s ``` Test Plan: COMPILE_WITH_ASAN=1 make check -j64 Reviewers: yhchiang, andrewkr, sdong Reviewed By: sdong Subscribers: lovro, andrewkr, dhruba Differential Revision: https://reviews.facebook.net/D57075
2016-07-20 16:49:03 +00:00
merge_context_.Clear();
keys_.emplace_front(std::move(original_key));
Introduce FullMergeV2 (eliminate memcpy from merge operators) Summary: This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice> This diff is stacked on top of D56493 and D56511 In this diff we - Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future - Replace std::deque<std::string> with std::vector<Slice> to pass operands - Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187) - Allow FullMergeV2 output to be an existing operand ``` [Everything in Memtable | 10K operands | 10 KB each | 1 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s [master] readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s ``` ``` [Everything in Memtable | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s [master] readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 1 operand per key] [FullMergeV2] $ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s [master] readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions [FullMergeV2] readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s [master] readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s ``` Test Plan: COMPILE_WITH_ASAN=1 make check -j64 Reviewers: yhchiang, andrewkr, sdong Reviewed By: sdong Subscribers: lovro, andrewkr, dhruba Differential Revision: https://reviews.facebook.net/D57075
2016-07-20 16:49:03 +00:00
merge_context_.PushOperand(merge_result);
} else if (op_failure_scope == MergeOperator::OpFailureScope::kMustMerge) {
// Change to `Status::MergeInProgress()` to denote output consists of
// merge operands only.
s = Status::MergeInProgress();
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences. Summary: Here are the major changes to the Merge Interface. It has been expanded to handle cases where the MergeOperator is not associative. It does so by stacking up merge operations while scanning through the key history (i.e.: during Get() or Compaction), until a valid Put/Delete/end-of-history is encountered; it then applies all of the merge operations in the correct sequence starting with the base/sentinel value. I have also introduced an "AssociativeMerge" function which allows the user to take advantage of associative merge operations (such as in the case of counters). The implementation will always attempt to merge the operations/operands themselves together when they are encountered, and will resort to the "stacking" method if and only if the "associative-merge" fails. This implementation is conjectured to allow MergeOperator to handle the general case, while still providing the user with the ability to take advantage of certain efficiencies in their own merge-operator / data-structure. NOTE: This is a preliminary diff. This must still go through a lot of review, revision, and testing. Feedback welcome! Test Plan: -This is a preliminary diff. I have only just begun testing/debugging it. -I will be testing this with the existing MergeOperator use-cases and unit-tests (counters, string-append, and redis-lists) -I will be "desk-checking" and walking through the code with the help gdb. -I will find a way of stress-testing the new interface / implementation using db_bench, db_test, merge_test, and/or db_stress. -I will ensure that my tests cover all cases: Get-Memtable, Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0, Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found, end-of-history, end-of-file, etc. -A lot of feedback from the reviewers. Reviewers: haobo, dhruba, zshao, emayanke Reviewed By: haobo CC: leveldb Differential Revision: https://reviews.facebook.net/D11499
2013-08-06 03:14:32 +00:00
}
} else {
// We haven't seen the beginning of the key nor a Put/Delete.
// Attempt to use the user's associative merge function to
// merge the stacked merge operands into a single operand.
s = Status::MergeInProgress();
if (merge_context_.GetNumOperands() >= 2 ||
(allow_single_operand_ && merge_context_.GetNumOperands() == 1)) {
bool merge_success = false;
std::string merge_result;
{
StopWatchNano timer(clock_, stats_ != nullptr);
PERF_TIMER_GUARD(merge_operator_time_nanos);
merge_success = user_merge_operator_->PartialMergeMulti(
orig_ikey.user_key,
Introduce FullMergeV2 (eliminate memcpy from merge operators) Summary: This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice> This diff is stacked on top of D56493 and D56511 In this diff we - Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future - Replace std::deque<std::string> with std::vector<Slice> to pass operands - Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187) - Allow FullMergeV2 output to be an existing operand ``` [Everything in Memtable | 10K operands | 10 KB each | 1 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s [master] readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s ``` ``` [Everything in Memtable | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s [master] readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 1 operand per key] [FullMergeV2] $ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s [master] readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions [FullMergeV2] readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s [master] readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s ``` Test Plan: COMPILE_WITH_ASAN=1 make check -j64 Reviewers: yhchiang, andrewkr, sdong Reviewed By: sdong Subscribers: lovro, andrewkr, dhruba Differential Revision: https://reviews.facebook.net/D57075
2016-07-20 16:49:03 +00:00
std::deque<Slice>(merge_context_.GetOperands().begin(),
merge_context_.GetOperands().end()),
&merge_result, logger_);
RecordTick(stats_, MERGE_OPERATION_TOTAL_TIME,
Remove wasteful instrumentation in FullMerge (stacked on D59577) Summary: [ This diff is stacked on top of D59577 ] We keep calling timer.ElapsedNanos() on every call to MergeOperator::FullMerge even when statistics are disabled, this is wasteful. I run the readseq benchmark on a DB containing 100K merge operands for 100K keys (1 operand per key) with 1GB block cache I see slight performance improvment Original results ``` $ ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=100000 --num=100000 --db="/dev/shm/100K_merge_compacted/" --cache_size=1073741824 --use_existing_db --disable_auto_compactions ------------------------------------------------ DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.498 micros/op 2006597 ops/sec; 222.0 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.295 micros/op 3393627 ops/sec; 375.4 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.285 micros/op 3511155 ops/sec; 388.4 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.286 micros/op 3500470 ops/sec; 387.2 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.283 micros/op 3530751 ops/sec; 390.6 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.289 micros/op 3464811 ops/sec; 383.3 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.277 micros/op 3612814 ops/sec; 399.7 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.283 micros/op 3539640 ops/sec; 391.6 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.285 micros/op 3503766 ops/sec; 387.6 MB/s ``` After patch ``` $ ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=100000 --num=100000 --db="/dev/shm/100K_merge_compacted/" --cache_size=1073741824 --use_existing_db --disable_auto_compactions ------------------------------------------------ DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.476 micros/op 2100119 ops/sec; 232.3 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.278 micros/op 3600887 ops/sec; 398.4 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.275 micros/op 3636698 ops/sec; 402.3 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.271 micros/op 3691661 ops/sec; 408.4 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.273 micros/op 3661534 ops/sec; 405.1 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.276 micros/op 3627106 ops/sec; 401.3 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.272 micros/op 3682635 ops/sec; 407.4 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.266 micros/op 3758331 ops/sec; 415.8 MB/s DB path: [/dev/shm/100K_merge_compacted/] readseq : 0.266 micros/op 3761907 ops/sec; 416.2 MB/s ``` Test Plan: make check -j64 Reviewers: yhchiang, sdong Reviewed By: sdong Subscribers: andrewkr, dhruba Differential Revision: https://reviews.facebook.net/D59583
2016-06-13 23:22:14 +00:00
stats_ ? timer.ElapsedNanosSafe() : 0);
}
if (merge_success) {
// Merging of operands (associative merge) was successful.
// Replace operands with the merge result
Introduce FullMergeV2 (eliminate memcpy from merge operators) Summary: This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice> This diff is stacked on top of D56493 and D56511 In this diff we - Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future - Replace std::deque<std::string> with std::vector<Slice> to pass operands - Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187) - Allow FullMergeV2 output to be an existing operand ``` [Everything in Memtable | 10K operands | 10 KB each | 1 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s [master] readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s ``` ``` [Everything in Memtable | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s [master] readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 1 operand per key] [FullMergeV2] $ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s [master] readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions [FullMergeV2] readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s [master] readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s ``` Test Plan: COMPILE_WITH_ASAN=1 make check -j64 Reviewers: yhchiang, andrewkr, sdong Reviewed By: sdong Subscribers: lovro, andrewkr, dhruba Differential Revision: https://reviews.facebook.net/D57075
2016-07-20 16:49:03 +00:00
merge_context_.Clear();
merge_context_.PushOperand(merge_result);
keys_.erase(keys_.begin(), keys_.end() - 1);
}
}
}
return s;
}
MergeOutputIterator::MergeOutputIterator(const MergeHelper* merge_helper)
: merge_helper_(merge_helper) {
it_keys_ = merge_helper_->keys().rend();
it_values_ = merge_helper_->values().rend();
}
void MergeOutputIterator::SeekToFirst() {
const auto& keys = merge_helper_->keys();
const auto& values = merge_helper_->values();
assert(keys.size() == values.size());
it_keys_ = keys.rbegin();
it_values_ = values.rbegin();
}
void MergeOutputIterator::Next() {
++it_keys_;
++it_values_;
}
CompactionFilter::Decision MergeHelper::FilterMerge(const Slice& user_key,
const Slice& value_slice) {
if (compaction_filter_ == nullptr) {
return CompactionFilter::Decision::kKeep;
}
if (stats_ != nullptr && ShouldReportDetailedTime(env_, stats_)) {
filter_timer_.Start();
}
compaction_filter_value_.clear();
compaction_filter_skip_until_.Clear();
auto ret = compaction_filter_->FilterV3(
level_, user_key, CompactionFilter::ValueType::kMergeOperand,
&value_slice, /* existing_columns */ nullptr, &compaction_filter_value_,
/* new_columns */ nullptr, compaction_filter_skip_until_.rep());
if (ret == CompactionFilter::Decision::kRemoveAndSkipUntil) {
if (user_comparator_->Compare(*compaction_filter_skip_until_.rep(),
user_key) <= 0) {
// Invalid skip_until returned from compaction filter.
// Keep the key as per FilterV2/FilterV3 documentation.
ret = CompactionFilter::Decision::kKeep;
} else {
compaction_filter_skip_until_.ConvertFromUserKey(kMaxSequenceNumber,
kValueTypeForSeek);
}
}
if (stats_ != nullptr && ShouldReportDetailedTime(env_, stats_)) {
total_filter_time_ += filter_timer_.ElapsedNanosSafe();
}
return ret;
}
} // namespace ROCKSDB_NAMESPACE