rocksdb/db/table_cache_sync_and_async.h

136 lines
5 KiB
C
Raw Normal View History

// Copyright (c) Meta Platforms, Inc. and affiliates.
//
Multi file concurrency in MultiGet using coroutines and async IO (#9968) Summary: This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code. A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest. TODO: 1. Figure out how to build it in CircleCI (requires some dependencies to be installed) 2. Do some stress testing with coroutines enabled No regression in synchronous MultiGet between this branch and main - ``` ./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics ``` Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)``` Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)``` More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file. 1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) - No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)``` Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)``` 2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file - No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)``` Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)``` 3. Single thread CPU bound workload with ~2 key overlap/file - No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)``` Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)``` 4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file - No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ``` Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968 Reviewed By: akankshamahajan15 Differential Revision: D36348563 Pulled By: anand1976 fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
2022-05-19 22:36:27 +00:00
// 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 "util/coro_utils.h"
#if defined(WITHOUT_COROUTINES) || \
(defined(USE_COROUTINES) && defined(WITH_COROUTINES))
namespace ROCKSDB_NAMESPACE {
#if defined(WITHOUT_COROUTINES)
#endif
// Batched version of TableCache::MultiGet.
DEFINE_SYNC_AND_ASYNC(Status, TableCache::MultiGet)
(const ReadOptions& options, const InternalKeyComparator& internal_comparator,
const FileMetaData& file_meta, const MultiGetContext::Range* mget_range,
Block per key-value checksum (#11287) Summary: add option `block_protection_bytes_per_key` and implementation for block per key-value checksum. The main changes are 1. checksum construction and verification in block.cc/h 2. pass the option `block_protection_bytes_per_key` around (mainly for methods defined in table_cache.h) 3. unit tests/crash test updates Tests: * Added unit tests * Crash test: `python3 tools/db_crashtest.py blackbox --simple --block_protection_bytes_per_key=1 --write_buffer_size=1048576` Follow up (maybe as a separate PR): make sure corruption status returned from BlockIters are correctly handled. Performance: Turning on block per KV protection has a non-trivial negative impact on read performance and costs additional memory. For memory, each block includes additional 24 bytes for checksum-related states beside checksum itself. For CPU, I set up a DB of size ~1.2GB with 5M keys (32 bytes key and 200 bytes value) which compacts to ~5 SST files (target file size 256 MB) in L6 without compression. I tested readrandom performance with various block cache size (to mimic various cache hit rates): ``` SETUP make OPTIMIZE_LEVEL="-O3" USE_LTO=1 DEBUG_LEVEL=0 -j32 db_bench ./db_bench -benchmarks=fillseq,compact0,waitforcompaction,compact,waitforcompaction -write_buffer_size=33554432 -level_compaction_dynamic_level_bytes=true -max_background_jobs=8 -target_file_size_base=268435456 --num=5000000 --key_size=32 --value_size=200 --compression_type=none BENCHMARK ./db_bench --use_existing_db -benchmarks=readtocache,readrandom[-X10] --num=5000000 --key_size=32 --disable_auto_compactions --reads=1000000 --block_protection_bytes_per_key=[0|1] --cache_size=$CACHESIZE The readrandom ops/sec looks like the following: Block cache size: 2GB 1.2GB * 0.9 1.2GB * 0.8 1.2GB * 0.5 8MB Main 240805 223604 198176 161653 139040 PR prot_bytes=0 238691 226693 200127 161082 141153 PR prot_bytes=1 214983 193199 178532 137013 108211 prot_bytes=1 vs -10% -15% -10.8% -15% -23% prot_bytes=0 ``` The benchmark has a lot of variance, but there was a 5% to 25% regression in this benchmark with different cache hit rates. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11287 Reviewed By: ajkr Differential Revision: D43970708 Pulled By: cbi42 fbshipit-source-id: ef98d898b71779846fa74212b9ec9e08b7183940
2023-04-25 19:08:23 +00:00
uint8_t block_protection_bytes_per_key,
Multi file concurrency in MultiGet using coroutines and async IO (#9968) Summary: This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code. A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest. TODO: 1. Figure out how to build it in CircleCI (requires some dependencies to be installed) 2. Do some stress testing with coroutines enabled No regression in synchronous MultiGet between this branch and main - ``` ./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics ``` Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)``` Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)``` More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file. 1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) - No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)``` Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)``` 2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file - No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)``` Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)``` 3. Single thread CPU bound workload with ~2 key overlap/file - No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)``` Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)``` 4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file - No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ``` Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968 Reviewed By: akankshamahajan15 Differential Revision: D36348563 Pulled By: anand1976 fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
2022-05-19 22:36:27 +00:00
const std::shared_ptr<const SliceTransform>& prefix_extractor,
HistogramImpl* file_read_hist, bool skip_filters, bool skip_range_deletions,
Major Cache refactoring, CPU efficiency improvement (#10975) Summary: This is several refactorings bundled into one to avoid having to incrementally re-modify uses of Cache several times. Overall, there are breaking changes to Cache class, and it becomes more of low-level interface for implementing caches, especially block cache. New internal APIs make using Cache cleaner than before, and more insulated from block cache evolution. Hopefully, this is the last really big block cache refactoring, because of rather effectively decoupling the implementations from the uses. This change also removes the EXPERIMENTAL designation on the SecondaryCache support in Cache. It seems reasonably mature at this point but still subject to change/evolution (as I warn in the API docs for Cache). The high-level motivation for this refactoring is to minimize code duplication / compounding complexity in adding SecondaryCache support to HyperClockCache (in a later PR). Other benefits listed below. * static_cast lines of code +29 -35 (net removed 6) * reinterpret_cast lines of code +6 -32 (net removed 26) ## cache.h and secondary_cache.h * Always use CacheItemHelper with entries instead of just a Deleter. There are several motivations / justifications: * Simpler for implementations to deal with just one Insert and one Lookup. * Simpler and more efficient implementation because we don't have to track which entries are using helpers and which are using deleters * Gets rid of hack to classify cache entries by their deleter. Instead, the CacheItemHelper includes a CacheEntryRole. This simplifies a lot of code (cache_entry_roles.h almost eliminated). Fixes https://github.com/facebook/rocksdb/issues/9428. * Makes it trivial to adjust SecondaryCache behavior based on kind of block (e.g. don't re-compress filter blocks). * It is arguably less convenient for many direct users of Cache, but direct users of Cache are now rare with introduction of typed_cache.h (below). * I considered and rejected an alternative approach in which we reduce customizability by assuming each secondary cache compatible value starts with a Slice referencing the uncompressed block contents (already true or mostly true), but we apparently intend to stack secondary caches. Saving an entry from a compressed secondary to a lower tier requires custom handling offered by SaveToCallback, etc. * Make CreateCallback part of the helper and introduce CreateContext to work with it (alternative to https://github.com/facebook/rocksdb/issues/10562). This cleans up the interface while still allowing context to be provided for loading/parsing values into primary cache. This model works for async lookup in BlockBasedTable reader (reader owns a CreateContext) under the assumption that it always waits on secondary cache operations to finish. (Otherwise, the CreateContext could be destroyed while async operation depending on it continues.) This likely contributes most to the observed performance improvement because it saves an std::function backed by a heap allocation. * Use char* for serialized data, e.g. in SaveToCallback, where void* was confusingly used. (We use `char*` for serialized byte data all over RocksDB, with many advantages over `void*`. `memcpy` etc. are legacy APIs that should not be mimicked.) * Add a type alias Cache::ObjectPtr = void*, so that we can better indicate the intent of the void* when it is to be the object associated with a Cache entry. Related: started (but did not complete) a refactoring to move away from "value" of a cache entry toward "object" or "obj". (It is confusing to call Cache a key-value store (like DB) when it is really storing arbitrary in-memory objects, not byte strings.) * Remove unnecessary key param from DeleterFn. This is good for efficiency in HyperClockCache, which does not directly store the cache key in memory. (Alternative to https://github.com/facebook/rocksdb/issues/10774) * Add allocator to Cache DeleterFn. This is a kind of future-proofing change in case we get more serious about using the Cache allocator for memory tracked by the Cache. Right now, only the uncompressed block contents are allocated using the allocator, and a pointer to that allocator is saved as part of the cached object so that the deleter can use it. (See CacheAllocationPtr.) If in the future we are able to "flatten out" our Cache objects some more, it would be good not to have to track the allocator as part of each object. * Removes legacy `ApplyToAllCacheEntries` and changes `ApplyToAllEntries` signature for Deleter->CacheItemHelper change. ## typed_cache.h Adds various "typed" interfaces to the Cache as internal APIs, so that most uses of Cache can use simple type safe code without casting and without explicit deleters, etc. Almost all of the non-test, non-glue code uses of Cache have been migrated. (Follow-up work: CompressedSecondaryCache deserves deeper attention to migrate.) This change expands RocksDB's internal usage of metaprogramming and SFINAE (https://en.cppreference.com/w/cpp/language/sfinae). The existing usages of Cache are divided up at a high level into these new interfaces. See updated existing uses of Cache for examples of how these are used. * PlaceholderCacheInterface - Used for making cache reservations, with entries that have a charge but no value. * BasicTypedCacheInterface<TValue> - Used for primary cache storage of objects of type TValue, which can be cleaned up with std::default_delete<TValue>. The role is provided by TValue::kCacheEntryRole or given in an optional template parameter. * FullTypedCacheInterface<TValue, TCreateContext> - Used for secondary cache compatible storage of objects of type TValue. In addition to BasicTypedCacheInterface constraints, we require TValue::ContentSlice() to return persistable data. This simplifies usage for the normal case of simple secondary cache compatibility (can give you a Slice to the data already in memory). In addition to TCreateContext performing the role of Cache::CreateContext, it is also expected to provide a factory function for creating TValue. * For each of these, there's a "Shared" version (e.g. FullTypedSharedCacheInterface) that holds a shared_ptr to the Cache, rather than assuming external ownership by holding only a raw `Cache*`. These interfaces introduce specific handle types for each interface instantiation, so that it's easy to see what kind of object is controlled by a handle. (Ultimately, this might not be worth the extra complexity, but it seems OK so far.) Note: I attempted to make the cache 'charge' automatically inferred from the cache object type, such as by expecting an ApproximateMemoryUsage() function, but this is not so clean because there are cases where we need to compute the charge ahead of time and don't want to re-compute it. ## block_cache.h This header is essentially the replacement for the old block_like_traits.h. It includes various things to support block cache access with typed_cache.h for block-based table. ## block_based_table_reader.cc Before this change, accessing the block cache here was an awkward mix of static polymorphism (template TBlocklike) and switch-case on a dynamic BlockType value. This change mostly unifies on static polymorphism, relying on minor hacks in block_cache.h to distinguish variants of Block. We still check BlockType in some places (especially for stats, which could be improved in follow-up work) but at least the BlockType is a static constant from the template parameter. (No more awkward partial redundancy between static and dynamic info.) This likely contributes to the overall performance improvement, but hasn't been tested in isolation. The other key source of simplification here is a more unified system of creating block cache objects: for directly populating from primary cache and for promotion from secondary cache. Both use BlockCreateContext, for context and for factory functions. ## block_based_table_builder.cc, cache_dump_load_impl.cc Before this change, warming caches was super ugly code. Both of these source files had switch statements to basically transition from the dynamic BlockType world to the static TBlocklike world. None of that mess is needed anymore as there's a new, untyped WarmInCache function that handles all the details just as promotion from SecondaryCache would. (Fixes `TODO akanksha: Dedup below code` in block_based_table_builder.cc.) ## Everything else Mostly just updating Cache users to use new typed APIs when reasonably possible, or changed Cache APIs when not. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10975 Test Plan: tests updated Performance test setup similar to https://github.com/facebook/rocksdb/issues/10626 (by cache size, LRUCache when not "hyper" for HyperClockCache): 34MB 1thread base.hyper -> kops/s: 0.745 io_bytes/op: 2.52504e+06 miss_ratio: 0.140906 max_rss_mb: 76.4844 34MB 1thread new.hyper -> kops/s: 0.751 io_bytes/op: 2.5123e+06 miss_ratio: 0.140161 max_rss_mb: 79.3594 34MB 1thread base -> kops/s: 0.254 io_bytes/op: 1.36073e+07 miss_ratio: 0.918818 max_rss_mb: 45.9297 34MB 1thread new -> kops/s: 0.252 io_bytes/op: 1.36157e+07 miss_ratio: 0.918999 max_rss_mb: 44.1523 34MB 32thread base.hyper -> kops/s: 7.272 io_bytes/op: 2.88323e+06 miss_ratio: 0.162532 max_rss_mb: 516.602 34MB 32thread new.hyper -> kops/s: 7.214 io_bytes/op: 2.99046e+06 miss_ratio: 0.168818 max_rss_mb: 518.293 34MB 32thread base -> kops/s: 3.528 io_bytes/op: 1.35722e+07 miss_ratio: 0.914691 max_rss_mb: 264.926 34MB 32thread new -> kops/s: 3.604 io_bytes/op: 1.35744e+07 miss_ratio: 0.915054 max_rss_mb: 264.488 233MB 1thread base.hyper -> kops/s: 53.909 io_bytes/op: 2552.35 miss_ratio: 0.0440566 max_rss_mb: 241.984 233MB 1thread new.hyper -> kops/s: 62.792 io_bytes/op: 2549.79 miss_ratio: 0.044043 max_rss_mb: 241.922 233MB 1thread base -> kops/s: 1.197 io_bytes/op: 2.75173e+06 miss_ratio: 0.103093 max_rss_mb: 241.559 233MB 1thread new -> kops/s: 1.199 io_bytes/op: 2.73723e+06 miss_ratio: 0.10305 max_rss_mb: 240.93 233MB 32thread base.hyper -> kops/s: 1298.69 io_bytes/op: 2539.12 miss_ratio: 0.0440307 max_rss_mb: 371.418 233MB 32thread new.hyper -> kops/s: 1421.35 io_bytes/op: 2538.75 miss_ratio: 0.0440307 max_rss_mb: 347.273 233MB 32thread base -> kops/s: 9.693 io_bytes/op: 2.77304e+06 miss_ratio: 0.103745 max_rss_mb: 569.691 233MB 32thread new -> kops/s: 9.75 io_bytes/op: 2.77559e+06 miss_ratio: 0.103798 max_rss_mb: 552.82 1597MB 1thread base.hyper -> kops/s: 58.607 io_bytes/op: 1449.14 miss_ratio: 0.0249324 max_rss_mb: 1583.55 1597MB 1thread new.hyper -> kops/s: 69.6 io_bytes/op: 1434.89 miss_ratio: 0.0247167 max_rss_mb: 1584.02 1597MB 1thread base -> kops/s: 60.478 io_bytes/op: 1421.28 miss_ratio: 0.024452 max_rss_mb: 1589.45 1597MB 1thread new -> kops/s: 63.973 io_bytes/op: 1416.07 miss_ratio: 0.0243766 max_rss_mb: 1589.24 1597MB 32thread base.hyper -> kops/s: 1436.2 io_bytes/op: 1357.93 miss_ratio: 0.0235353 max_rss_mb: 1692.92 1597MB 32thread new.hyper -> kops/s: 1605.03 io_bytes/op: 1358.04 miss_ratio: 0.023538 max_rss_mb: 1702.78 1597MB 32thread base -> kops/s: 280.059 io_bytes/op: 1350.34 miss_ratio: 0.023289 max_rss_mb: 1675.36 1597MB 32thread new -> kops/s: 283.125 io_bytes/op: 1351.05 miss_ratio: 0.0232797 max_rss_mb: 1703.83 Almost uniformly improving over base revision, especially for hot paths with HyperClockCache, up to 12% higher throughput seen (1597MB, 32thread, hyper). The improvement for that is likely coming from much simplified code for providing context for secondary cache promotion (CreateCallback/CreateContext), and possibly from less branching in block_based_table_reader. And likely a small improvement from not reconstituting key for DeleterFn. Reviewed By: anand1976 Differential Revision: D42417818 Pulled By: pdillinger fbshipit-source-id: f86bfdd584dce27c028b151ba56818ad14f7a432
2023-01-11 22:20:40 +00:00
int level, TypedHandle* handle) {
Multi file concurrency in MultiGet using coroutines and async IO (#9968) Summary: This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code. A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest. TODO: 1. Figure out how to build it in CircleCI (requires some dependencies to be installed) 2. Do some stress testing with coroutines enabled No regression in synchronous MultiGet between this branch and main - ``` ./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics ``` Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)``` Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)``` More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file. 1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) - No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)``` Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)``` 2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file - No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)``` Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)``` 3. Single thread CPU bound workload with ~2 key overlap/file - No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)``` Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)``` 4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file - No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ``` Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968 Reviewed By: akankshamahajan15 Differential Revision: D36348563 Pulled By: anand1976 fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
2022-05-19 22:36:27 +00:00
auto& fd = file_meta.fd;
Status s;
TableReader* t = fd.table_reader;
MultiGetRange table_range(*mget_range, mget_range->begin(),
mget_range->end());
if (handle != nullptr && t == nullptr) {
Major Cache refactoring, CPU efficiency improvement (#10975) Summary: This is several refactorings bundled into one to avoid having to incrementally re-modify uses of Cache several times. Overall, there are breaking changes to Cache class, and it becomes more of low-level interface for implementing caches, especially block cache. New internal APIs make using Cache cleaner than before, and more insulated from block cache evolution. Hopefully, this is the last really big block cache refactoring, because of rather effectively decoupling the implementations from the uses. This change also removes the EXPERIMENTAL designation on the SecondaryCache support in Cache. It seems reasonably mature at this point but still subject to change/evolution (as I warn in the API docs for Cache). The high-level motivation for this refactoring is to minimize code duplication / compounding complexity in adding SecondaryCache support to HyperClockCache (in a later PR). Other benefits listed below. * static_cast lines of code +29 -35 (net removed 6) * reinterpret_cast lines of code +6 -32 (net removed 26) ## cache.h and secondary_cache.h * Always use CacheItemHelper with entries instead of just a Deleter. There are several motivations / justifications: * Simpler for implementations to deal with just one Insert and one Lookup. * Simpler and more efficient implementation because we don't have to track which entries are using helpers and which are using deleters * Gets rid of hack to classify cache entries by their deleter. Instead, the CacheItemHelper includes a CacheEntryRole. This simplifies a lot of code (cache_entry_roles.h almost eliminated). Fixes https://github.com/facebook/rocksdb/issues/9428. * Makes it trivial to adjust SecondaryCache behavior based on kind of block (e.g. don't re-compress filter blocks). * It is arguably less convenient for many direct users of Cache, but direct users of Cache are now rare with introduction of typed_cache.h (below). * I considered and rejected an alternative approach in which we reduce customizability by assuming each secondary cache compatible value starts with a Slice referencing the uncompressed block contents (already true or mostly true), but we apparently intend to stack secondary caches. Saving an entry from a compressed secondary to a lower tier requires custom handling offered by SaveToCallback, etc. * Make CreateCallback part of the helper and introduce CreateContext to work with it (alternative to https://github.com/facebook/rocksdb/issues/10562). This cleans up the interface while still allowing context to be provided for loading/parsing values into primary cache. This model works for async lookup in BlockBasedTable reader (reader owns a CreateContext) under the assumption that it always waits on secondary cache operations to finish. (Otherwise, the CreateContext could be destroyed while async operation depending on it continues.) This likely contributes most to the observed performance improvement because it saves an std::function backed by a heap allocation. * Use char* for serialized data, e.g. in SaveToCallback, where void* was confusingly used. (We use `char*` for serialized byte data all over RocksDB, with many advantages over `void*`. `memcpy` etc. are legacy APIs that should not be mimicked.) * Add a type alias Cache::ObjectPtr = void*, so that we can better indicate the intent of the void* when it is to be the object associated with a Cache entry. Related: started (but did not complete) a refactoring to move away from "value" of a cache entry toward "object" or "obj". (It is confusing to call Cache a key-value store (like DB) when it is really storing arbitrary in-memory objects, not byte strings.) * Remove unnecessary key param from DeleterFn. This is good for efficiency in HyperClockCache, which does not directly store the cache key in memory. (Alternative to https://github.com/facebook/rocksdb/issues/10774) * Add allocator to Cache DeleterFn. This is a kind of future-proofing change in case we get more serious about using the Cache allocator for memory tracked by the Cache. Right now, only the uncompressed block contents are allocated using the allocator, and a pointer to that allocator is saved as part of the cached object so that the deleter can use it. (See CacheAllocationPtr.) If in the future we are able to "flatten out" our Cache objects some more, it would be good not to have to track the allocator as part of each object. * Removes legacy `ApplyToAllCacheEntries` and changes `ApplyToAllEntries` signature for Deleter->CacheItemHelper change. ## typed_cache.h Adds various "typed" interfaces to the Cache as internal APIs, so that most uses of Cache can use simple type safe code without casting and without explicit deleters, etc. Almost all of the non-test, non-glue code uses of Cache have been migrated. (Follow-up work: CompressedSecondaryCache deserves deeper attention to migrate.) This change expands RocksDB's internal usage of metaprogramming and SFINAE (https://en.cppreference.com/w/cpp/language/sfinae). The existing usages of Cache are divided up at a high level into these new interfaces. See updated existing uses of Cache for examples of how these are used. * PlaceholderCacheInterface - Used for making cache reservations, with entries that have a charge but no value. * BasicTypedCacheInterface<TValue> - Used for primary cache storage of objects of type TValue, which can be cleaned up with std::default_delete<TValue>. The role is provided by TValue::kCacheEntryRole or given in an optional template parameter. * FullTypedCacheInterface<TValue, TCreateContext> - Used for secondary cache compatible storage of objects of type TValue. In addition to BasicTypedCacheInterface constraints, we require TValue::ContentSlice() to return persistable data. This simplifies usage for the normal case of simple secondary cache compatibility (can give you a Slice to the data already in memory). In addition to TCreateContext performing the role of Cache::CreateContext, it is also expected to provide a factory function for creating TValue. * For each of these, there's a "Shared" version (e.g. FullTypedSharedCacheInterface) that holds a shared_ptr to the Cache, rather than assuming external ownership by holding only a raw `Cache*`. These interfaces introduce specific handle types for each interface instantiation, so that it's easy to see what kind of object is controlled by a handle. (Ultimately, this might not be worth the extra complexity, but it seems OK so far.) Note: I attempted to make the cache 'charge' automatically inferred from the cache object type, such as by expecting an ApproximateMemoryUsage() function, but this is not so clean because there are cases where we need to compute the charge ahead of time and don't want to re-compute it. ## block_cache.h This header is essentially the replacement for the old block_like_traits.h. It includes various things to support block cache access with typed_cache.h for block-based table. ## block_based_table_reader.cc Before this change, accessing the block cache here was an awkward mix of static polymorphism (template TBlocklike) and switch-case on a dynamic BlockType value. This change mostly unifies on static polymorphism, relying on minor hacks in block_cache.h to distinguish variants of Block. We still check BlockType in some places (especially for stats, which could be improved in follow-up work) but at least the BlockType is a static constant from the template parameter. (No more awkward partial redundancy between static and dynamic info.) This likely contributes to the overall performance improvement, but hasn't been tested in isolation. The other key source of simplification here is a more unified system of creating block cache objects: for directly populating from primary cache and for promotion from secondary cache. Both use BlockCreateContext, for context and for factory functions. ## block_based_table_builder.cc, cache_dump_load_impl.cc Before this change, warming caches was super ugly code. Both of these source files had switch statements to basically transition from the dynamic BlockType world to the static TBlocklike world. None of that mess is needed anymore as there's a new, untyped WarmInCache function that handles all the details just as promotion from SecondaryCache would. (Fixes `TODO akanksha: Dedup below code` in block_based_table_builder.cc.) ## Everything else Mostly just updating Cache users to use new typed APIs when reasonably possible, or changed Cache APIs when not. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10975 Test Plan: tests updated Performance test setup similar to https://github.com/facebook/rocksdb/issues/10626 (by cache size, LRUCache when not "hyper" for HyperClockCache): 34MB 1thread base.hyper -> kops/s: 0.745 io_bytes/op: 2.52504e+06 miss_ratio: 0.140906 max_rss_mb: 76.4844 34MB 1thread new.hyper -> kops/s: 0.751 io_bytes/op: 2.5123e+06 miss_ratio: 0.140161 max_rss_mb: 79.3594 34MB 1thread base -> kops/s: 0.254 io_bytes/op: 1.36073e+07 miss_ratio: 0.918818 max_rss_mb: 45.9297 34MB 1thread new -> kops/s: 0.252 io_bytes/op: 1.36157e+07 miss_ratio: 0.918999 max_rss_mb: 44.1523 34MB 32thread base.hyper -> kops/s: 7.272 io_bytes/op: 2.88323e+06 miss_ratio: 0.162532 max_rss_mb: 516.602 34MB 32thread new.hyper -> kops/s: 7.214 io_bytes/op: 2.99046e+06 miss_ratio: 0.168818 max_rss_mb: 518.293 34MB 32thread base -> kops/s: 3.528 io_bytes/op: 1.35722e+07 miss_ratio: 0.914691 max_rss_mb: 264.926 34MB 32thread new -> kops/s: 3.604 io_bytes/op: 1.35744e+07 miss_ratio: 0.915054 max_rss_mb: 264.488 233MB 1thread base.hyper -> kops/s: 53.909 io_bytes/op: 2552.35 miss_ratio: 0.0440566 max_rss_mb: 241.984 233MB 1thread new.hyper -> kops/s: 62.792 io_bytes/op: 2549.79 miss_ratio: 0.044043 max_rss_mb: 241.922 233MB 1thread base -> kops/s: 1.197 io_bytes/op: 2.75173e+06 miss_ratio: 0.103093 max_rss_mb: 241.559 233MB 1thread new -> kops/s: 1.199 io_bytes/op: 2.73723e+06 miss_ratio: 0.10305 max_rss_mb: 240.93 233MB 32thread base.hyper -> kops/s: 1298.69 io_bytes/op: 2539.12 miss_ratio: 0.0440307 max_rss_mb: 371.418 233MB 32thread new.hyper -> kops/s: 1421.35 io_bytes/op: 2538.75 miss_ratio: 0.0440307 max_rss_mb: 347.273 233MB 32thread base -> kops/s: 9.693 io_bytes/op: 2.77304e+06 miss_ratio: 0.103745 max_rss_mb: 569.691 233MB 32thread new -> kops/s: 9.75 io_bytes/op: 2.77559e+06 miss_ratio: 0.103798 max_rss_mb: 552.82 1597MB 1thread base.hyper -> kops/s: 58.607 io_bytes/op: 1449.14 miss_ratio: 0.0249324 max_rss_mb: 1583.55 1597MB 1thread new.hyper -> kops/s: 69.6 io_bytes/op: 1434.89 miss_ratio: 0.0247167 max_rss_mb: 1584.02 1597MB 1thread base -> kops/s: 60.478 io_bytes/op: 1421.28 miss_ratio: 0.024452 max_rss_mb: 1589.45 1597MB 1thread new -> kops/s: 63.973 io_bytes/op: 1416.07 miss_ratio: 0.0243766 max_rss_mb: 1589.24 1597MB 32thread base.hyper -> kops/s: 1436.2 io_bytes/op: 1357.93 miss_ratio: 0.0235353 max_rss_mb: 1692.92 1597MB 32thread new.hyper -> kops/s: 1605.03 io_bytes/op: 1358.04 miss_ratio: 0.023538 max_rss_mb: 1702.78 1597MB 32thread base -> kops/s: 280.059 io_bytes/op: 1350.34 miss_ratio: 0.023289 max_rss_mb: 1675.36 1597MB 32thread new -> kops/s: 283.125 io_bytes/op: 1351.05 miss_ratio: 0.0232797 max_rss_mb: 1703.83 Almost uniformly improving over base revision, especially for hot paths with HyperClockCache, up to 12% higher throughput seen (1597MB, 32thread, hyper). The improvement for that is likely coming from much simplified code for providing context for secondary cache promotion (CreateCallback/CreateContext), and possibly from less branching in block_based_table_reader. And likely a small improvement from not reconstituting key for DeleterFn. Reviewed By: anand1976 Differential Revision: D42417818 Pulled By: pdillinger fbshipit-source-id: f86bfdd584dce27c028b151ba56818ad14f7a432
2023-01-11 22:20:40 +00:00
t = cache_.Value(handle);
}
Multi file concurrency in MultiGet using coroutines and async IO (#9968) Summary: This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code. A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest. TODO: 1. Figure out how to build it in CircleCI (requires some dependencies to be installed) 2. Do some stress testing with coroutines enabled No regression in synchronous MultiGet between this branch and main - ``` ./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics ``` Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)``` Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)``` More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file. 1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) - No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)``` Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)``` 2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file - No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)``` Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)``` 3. Single thread CPU bound workload with ~2 key overlap/file - No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)``` Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)``` 4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file - No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ``` Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968 Reviewed By: akankshamahajan15 Differential Revision: D36348563 Pulled By: anand1976 fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
2022-05-19 22:36:27 +00:00
autovector<std::string, MultiGetContext::MAX_BATCH_SIZE> row_cache_entries;
IterKey row_cache_key;
size_t row_cache_key_prefix_size = 0;
KeyContext& first_key = *table_range.begin();
bool lookup_row_cache =
ioptions_.row_cache && !first_key.get_context->NeedToReadSequence();
// Check row cache if enabled. Since row cache does not currently store
// sequence numbers, we cannot use it if we need to fetch the sequence.
if (lookup_row_cache) {
GetContext* first_context = first_key.get_context;
CreateRowCacheKeyPrefix(options, fd, first_key.ikey, first_context,
row_cache_key);
row_cache_key_prefix_size = row_cache_key.Size();
for (auto miter = table_range.begin(); miter != table_range.end();
++miter) {
const Slice& user_key = miter->ukey_with_ts;
GetContext* get_context = miter->get_context;
Fix kBlockCacheTier read when merge-chain base value is in a blob file (#12462) Summary: The original goal is to propagate failures from `GetContext::SaveValue()` -> `GetContext::GetBlobValue()` -> `BlobFetcher::FetchBlob()` up to the user. This call sequence happens when a merge chain ends with a base value in a blob file. There's also fixes for bugs encountered along the way where non-ok statuses were ignored/overwritten, and a bit of plumbing work for functions that had no capability to return a status. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12462 Test Plan: A repro command ``` db=/dev/shm/dbstress_db ; exp=/dev/shm/dbstress_exp ; rm -rf $db $exp ; mkdir -p $db $exp ./db_stress \ --clear_column_family_one_in=0 \ --test_batches_snapshots=0 \ --write_fault_one_in=0 \ --use_put_entity_one_in=0 \ --prefixpercent=0 \ --read_fault_one_in=0 \ --readpercent=0 \ --reopen=0 \ --set_options_one_in=10000 \ --delpercent=0 \ --delrangepercent=0 \ --open_metadata_write_fault_one_in=0 \ --open_read_fault_one_in=0 \ --open_write_fault_one_in=0 \ --destroy_db_initially=0 \ --ingest_external_file_one_in=0 \ --iterpercent=0 \ --nooverwritepercent=0 \ --db=$db \ --enable_blob_files=1 \ --expected_values_dir=$exp \ --max_background_compactions=20 \ --max_bytes_for_level_base=2097152 \ --max_key=100000 \ --min_blob_size=0 \ --open_files=-1 \ --ops_per_thread=100000000 \ --prefix_size=-1 \ --target_file_size_base=524288 \ --use_merge=1 \ --value_size_mult=32 \ --write_buffer_size=524288 \ --writepercent=100 ``` It used to fail like: ``` ... frame https://github.com/facebook/rocksdb/issues/9: 0x00007fc63903bc93 libc.so.6`__GI___assert_fail(assertion="HasDefaultColumn(columns)", file="fbcode/internal_repo_rocksdb/repo/db/wide/wide_columns_helper.h", line=33, function="static const rocksdb::Slice &rocksdb::WideColumnsHelper::GetDefaultColumn(const rocksdb::WideColumns &)") at assert.c:101:3 frame https://github.com/facebook/rocksdb/issues/10: 0x00000000006f7e92 db_stress`rocksdb::Version::Get(rocksdb::ReadOptions const&, rocksdb::LookupKey const&, rocksdb::PinnableSlice*, rocksdb::PinnableWideColumns*, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>*, rocksdb::Status*, rocksdb::MergeContext*, unsigned long*, rocksdb::PinnedIteratorsManager*, bool*, bool*, unsigned long*, rocksdb::ReadCallback*, bool*, bool) [inlined] rocksdb::WideColumnsHelper::GetDefaultColumn(columns=size=0) at wide_columns_helper.h:33 frame https://github.com/facebook/rocksdb/issues/11: 0x00000000006f7e76 db_stress`rocksdb::Version::Get(this=0x00007fc5ec763000, read_options=<unavailable>, k=<unavailable>, value=0x0000000000000000, columns=0x00007fc6035fd1d8, timestamp=<unavailable>, status=0x00007fc6035fd250, merge_context=0x00007fc6035fce40, max_covering_tombstone_seq=0x00007fc6035fce90, pinned_iters_mgr=0x00007fc6035fcdf0, value_found=0x0000000000000000, key_exists=0x0000000000000000, seq=0x0000000000000000, callback=0x0000000000000000, is_blob=0x0000000000000000, do_merge=<unavailable>) at version_set.cc:2492 frame https://github.com/facebook/rocksdb/issues/12: 0x000000000051e245 db_stress`rocksdb::DBImpl::GetImpl(this=0x00007fc637a86000, read_options=0x00007fc6035fcf60, key=<unavailable>, get_impl_options=0x00007fc6035fd000) at db_impl.cc:2408 frame https://github.com/facebook/rocksdb/issues/13: 0x000000000050cec2 db_stress`rocksdb::DBImpl::GetEntity(this=0x00007fc637a86000, _read_options=<unavailable>, column_family=<unavailable>, key=0x00007fc6035fd3c8, columns=0x00007fc6035fd1d8) at db_impl.cc:2109 frame https://github.com/facebook/rocksdb/issues/14: 0x000000000074f688 db_stress`rocksdb::(anonymous namespace)::MemTableInserter::MergeCF(this=0x00007fc6035fd450, column_family_id=2, key=0x00007fc6035fd3c8, value=0x00007fc6035fd3a0) at write_batch.cc:2656 frame https://github.com/facebook/rocksdb/issues/15: 0x00000000007476fc db_stress`rocksdb::WriteBatchInternal::Iterate(wb=0x00007fc6035fe698, handler=0x00007fc6035fd450, begin=12, end=<unavailable>) at write_batch.cc:607 frame https://github.com/facebook/rocksdb/issues/16: 0x000000000074d7dd db_stress`rocksdb::WriteBatchInternal::InsertInto(rocksdb::WriteThread::WriteGroup&, unsigned long, rocksdb::ColumnFamilyMemTables*, rocksdb::FlushScheduler*, rocksdb::TrimHistoryScheduler*, bool, unsigned long, rocksdb::DB*, bool, bool, bool) [inlined] rocksdb::WriteBatch::Iterate(this=<unavailable>, handler=0x00007fc6035fd450) const at write_batch.cc:505 frame https://github.com/facebook/rocksdb/issues/17: 0x000000000074d77b db_stress`rocksdb::WriteBatchInternal::InsertInto(write_group=<unavailable>, sequence=<unavailable>, memtables=<unavailable>, flush_scheduler=<unavailable>, trim_history_scheduler=<unavailable>, ignore_missing_column_families=<unavailable>, recovery_log_number=0, db=0x00007fc637a86000, concurrent_memtable_writes=<unavailable>, seq_per_batch=false, batch_per_txn=<unavailable>) at write_batch.cc:3084 frame https://github.com/facebook/rocksdb/issues/18: 0x0000000000631d77 db_stress`rocksdb::DBImpl::PipelinedWriteImpl(this=0x00007fc637a86000, write_options=<unavailable>, my_batch=0x00007fc6035fe698, callback=0x0000000000000000, log_used=<unavailable>, log_ref=0, disable_memtable=<unavailable>, seq_used=0x0000000000000000) at db_impl_write.cc:807 frame https://github.com/facebook/rocksdb/issues/19: 0x000000000062ceeb db_stress`rocksdb::DBImpl::WriteImpl(this=<unavailable>, write_options=<unavailable>, my_batch=0x00007fc6035fe698, callback=0x0000000000000000, log_used=<unavailable>, log_ref=0, disable_memtable=<unavailable>, seq_used=0x0000000000000000, batch_cnt=0, pre_release_callback=0x0000000000000000, post_memtable_callback=0x0000000000000000) at db_impl_write.cc:312 frame https://github.com/facebook/rocksdb/issues/20: 0x000000000062c8ec db_stress`rocksdb::DBImpl::Write(this=0x00007fc637a86000, write_options=0x00007fc6035feca8, my_batch=0x00007fc6035fe698) at db_impl_write.cc:157 frame https://github.com/facebook/rocksdb/issues/21: 0x000000000062b847 db_stress`rocksdb::DB::Merge(this=0x00007fc637a86000, opt=0x00007fc6035feca8, column_family=0x00007fc6370bf140, key=0x00007fc6035fe8d8, value=0x00007fc6035fe830) at db_impl_write.cc:2544 frame https://github.com/facebook/rocksdb/issues/22: 0x000000000062b6ef db_stress`rocksdb::DBImpl::Merge(this=0x00007fc637a86000, o=<unavailable>, column_family=0x00007fc6370bf140, key=0x00007fc6035fe8d8, val=0x00007fc6035fe830) at db_impl_write.cc:72 frame https://github.com/facebook/rocksdb/issues/23: 0x00000000004d6397 db_stress`rocksdb::NonBatchedOpsStressTest::TestPut(this=0x00007fc637041000, thread=0x00007fc6370dbc00, write_opts=0x00007fc6035feca8, read_opts=0x00007fc6035fe9c8, rand_column_families=<unavailable>, rand_keys=size=1, value={P\xe9_\x03\xc6\x7f\0\0}) at no_batched_ops_stress.cc:1317 frame https://github.com/facebook/rocksdb/issues/24: 0x000000000049361d db_stress`rocksdb::StressTest::OperateDb(this=0x00007fc637041000, thread=0x00007fc6370dbc00) at db_stress_test_base.cc:1148 ... ``` Reviewed By: ltamasi Differential Revision: D55157795 Pulled By: ajkr fbshipit-source-id: 5f7c1380ead5794c29d41680028e34b839744764
2024-03-21 19:38:53 +00:00
Status read_status;
bool ret =
GetFromRowCache(user_key, row_cache_key, row_cache_key_prefix_size,
get_context, &read_status);
if (!read_status.ok()) {
CO_RETURN read_status;
}
if (ret) {
Multi file concurrency in MultiGet using coroutines and async IO (#9968) Summary: This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code. A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest. TODO: 1. Figure out how to build it in CircleCI (requires some dependencies to be installed) 2. Do some stress testing with coroutines enabled No regression in synchronous MultiGet between this branch and main - ``` ./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics ``` Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)``` Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)``` More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file. 1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) - No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)``` Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)``` 2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file - No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)``` Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)``` 3. Single thread CPU bound workload with ~2 key overlap/file - No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)``` Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)``` 4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file - No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ``` Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968 Reviewed By: akankshamahajan15 Differential Revision: D36348563 Pulled By: anand1976 fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
2022-05-19 22:36:27 +00:00
table_range.SkipKey(miter);
} else {
row_cache_entries.emplace_back();
get_context->SetReplayLog(&(row_cache_entries.back()));
}
}
}
// Check that table_range is not empty. Its possible all keys may have been
// found in the row cache and thus the range may now be empty
if (s.ok() && !table_range.empty()) {
if (t == nullptr) {
assert(handle == nullptr);
Always verify SST unique IDs on SST file open (#10532) Summary: Although we've been tracking SST unique IDs in the DB manifest unconditionally, checking has been opt-in and with an extra pass at DB::Open time. This changes the behavior of `verify_sst_unique_id_in_manifest` to check unique ID against manifest every time an SST file is opened through table cache (normal DB operations), replacing the explicit pass over files at DB::Open time. This change also enables the option by default and removes the "EXPERIMENTAL" designation. One possible criticism is that the option no longer ensures the integrity of a DB at Open time. This is far from an all-or-nothing issue. Verifying the IDs of all SST files hardly ensures all the data in the DB is readable. (VerifyChecksum is supposed to do that.) Also, with max_open_files=-1 (default, extremely common), all SST files are opened at DB::Open time anyway. Implementation details: * `VerifySstUniqueIdInManifest()` functions are the extra/explicit pass that is now removed. * Unit tests that manipulate/corrupt table properties have to opt out of this check, because that corrupts the "actual" unique id. (And even for testing we don't currently have a mechanism to set "no unique id" in the in-memory file metadata for new files.) * A lot of other unit test churn relates to (a) default checking on, and (b) checking on SST open even without DB::Open (e.g. on flush) * Use `FileMetaData` for more `TableCache` operations (in place of `FileDescriptor`) so that we have access to the unique_id whenever we might need to open an SST file. **There is the possibility of performance impact because we can no longer use the more localized `fd` part of an `FdWithKeyRange` but instead follow the `file_metadata` pointer. However, this change (possible regression) is only done for `GetMemoryUsageByTableReaders`.** * Removed a completely unnecessary constructor overload of `TableReaderOptions` Possible follow-up: * Verification only happens when opening through table cache. Are there more places where this should happen? * Improve error message when there is a file size mismatch vs. manifest (FIXME added in the appropriate place). * I'm not sure there's a justification for `FileDescriptor` to be distinct from `FileMetaData`. * I'm skeptical that `FdWithKeyRange` really still makes sense for optimizing some data locality by duplicating some data in memory, but I could be wrong. * An unnecessary overload of NewTableReader was recently added, in the public API nonetheless (though unusable there). It should be cleaned up to put most things under `TableReaderOptions`. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10532 Test Plan: updated unit tests Performance test showing no significant difference (just noise I think): `./db_bench -benchmarks=readwhilewriting[-X10] -num=3000000 -disable_wal=1 -bloom_bits=8 -write_buffer_size=1000000 -target_file_size_base=1000000` Before: readwhilewriting [AVG 10 runs] : 68702 (± 6932) ops/sec After: readwhilewriting [AVG 10 runs] : 68239 (± 7198) ops/sec Reviewed By: jay-zhuang Differential Revision: D38765551 Pulled By: pdillinger fbshipit-source-id: a827a708155f12344ab2a5c16e7701c7636da4c2
2022-09-08 05:52:42 +00:00
s = FindTable(options, file_options_, internal_comparator, file_meta,
Block per key-value checksum (#11287) Summary: add option `block_protection_bytes_per_key` and implementation for block per key-value checksum. The main changes are 1. checksum construction and verification in block.cc/h 2. pass the option `block_protection_bytes_per_key` around (mainly for methods defined in table_cache.h) 3. unit tests/crash test updates Tests: * Added unit tests * Crash test: `python3 tools/db_crashtest.py blackbox --simple --block_protection_bytes_per_key=1 --write_buffer_size=1048576` Follow up (maybe as a separate PR): make sure corruption status returned from BlockIters are correctly handled. Performance: Turning on block per KV protection has a non-trivial negative impact on read performance and costs additional memory. For memory, each block includes additional 24 bytes for checksum-related states beside checksum itself. For CPU, I set up a DB of size ~1.2GB with 5M keys (32 bytes key and 200 bytes value) which compacts to ~5 SST files (target file size 256 MB) in L6 without compression. I tested readrandom performance with various block cache size (to mimic various cache hit rates): ``` SETUP make OPTIMIZE_LEVEL="-O3" USE_LTO=1 DEBUG_LEVEL=0 -j32 db_bench ./db_bench -benchmarks=fillseq,compact0,waitforcompaction,compact,waitforcompaction -write_buffer_size=33554432 -level_compaction_dynamic_level_bytes=true -max_background_jobs=8 -target_file_size_base=268435456 --num=5000000 --key_size=32 --value_size=200 --compression_type=none BENCHMARK ./db_bench --use_existing_db -benchmarks=readtocache,readrandom[-X10] --num=5000000 --key_size=32 --disable_auto_compactions --reads=1000000 --block_protection_bytes_per_key=[0|1] --cache_size=$CACHESIZE The readrandom ops/sec looks like the following: Block cache size: 2GB 1.2GB * 0.9 1.2GB * 0.8 1.2GB * 0.5 8MB Main 240805 223604 198176 161653 139040 PR prot_bytes=0 238691 226693 200127 161082 141153 PR prot_bytes=1 214983 193199 178532 137013 108211 prot_bytes=1 vs -10% -15% -10.8% -15% -23% prot_bytes=0 ``` The benchmark has a lot of variance, but there was a 5% to 25% regression in this benchmark with different cache hit rates. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11287 Reviewed By: ajkr Differential Revision: D43970708 Pulled By: cbi42 fbshipit-source-id: ef98d898b71779846fa74212b9ec9e08b7183940
2023-04-25 19:08:23 +00:00
&handle, block_protection_bytes_per_key, prefix_extractor,
Multi file concurrency in MultiGet using coroutines and async IO (#9968) Summary: This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code. A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest. TODO: 1. Figure out how to build it in CircleCI (requires some dependencies to be installed) 2. Do some stress testing with coroutines enabled No regression in synchronous MultiGet between this branch and main - ``` ./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics ``` Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)``` Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)``` More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file. 1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) - No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)``` Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)``` 2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file - No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)``` Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)``` 3. Single thread CPU bound workload with ~2 key overlap/file - No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)``` Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)``` 4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file - No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ``` Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968 Reviewed By: akankshamahajan15 Differential Revision: D36348563 Pulled By: anand1976 fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
2022-05-19 22:36:27 +00:00
options.read_tier == kBlockCacheTier /* no_io */,
Fix StopWatch bug; Remove setting `record_read_stats` (#11474) Summary: **Context/Summary:** - StopWatch enable stats even when `StatsLevel::kExceptTimers` is set. It's a harmless bug though since `reportTimeToHistogram()` will not report it anyway according to https://github.com/facebook/rocksdb/blob/main/include/rocksdb/statistics.h#L705 - https://github.com/facebook/rocksdb/pull/11288 should have removed logics of setting `record_read_stats = !for_compaction` as we don't differentiate `RandomAccessFileReader`'s stats behavior based on compaction or not (instead we now report stats of different IO activities including compaction to different stats). Fixing this should report more compaction related file read micros that aren't reported previously due to `for_compaction==true` Pull Request resolved: https://github.com/facebook/rocksdb/pull/11474 Test Plan: - DB bench pre vs post fix with small max_open_files Setup command `./db_ bench -db=/dev/shm/testdb/ -statistics=true -benchmarks=fillseq -key_size=32 -value_size=512 -num=5000 -write_buffer_size=655 -target_file_size_base=655 -disable_auto_compactions=true -compression_type=none -bloom_bits=3` Run command `./db_bench --open_files=1 -use_existing_db=true -db=/dev/shm/testdb2/ -statistics=true -benchmarks=compactall -key_size=32 -value_size=512 -num=5000 -write_buffer_size=655 -target_file_size_base=655 -disable_auto_compactions=true -compression_type=none -bloom_bits=3` Pre-fix ``` rocksdb.sst.read.micros P50 : 2.056175 P95 : 4.647739 P99 : 8.948475 P100 : 25.000000 COUNT : 4451 SUM : 12827 rocksdb.file.read.flush.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0 rocksdb.file.read.compaction.micros P50 : 2.057397 P95 : 4.625253 P99 : 8.749474 P100 : 25.000000 COUNT : 4382 SUM : 12608 rocksdb.file.read.db.open.micros P50 : 1.985294 P95 : 9.100000 P99 : 13.000000 P100 : 13.000000 COUNT : 69 SUM : 219 ``` Post-fix (with a higher `rocksdb.file.read.compaction.micros` count) ``` rocksdb.sst.read.micros P50 : 1.858968 P95 : 3.653086 P99 : 5.968000 P100 : 21.000000 COUNT : 3548 SUM : 9119 rocksdb.file.read.flush.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0 rocksdb.file.read.compaction.micros P50 : 1.857027 P95 : 3.627614 P99 : 5.738621 P100 : 21.000000 COUNT : 3479 SUM : 8904 rocksdb.file.read.db.open.micros P50 : 2.000000 P95 : 6.733333 P99 : 11.000000 P100 : 11.000000 COUNT : 69 SUM : 215 ``` - CI Reviewed By: ajkr Differential Revision: D46137221 Pulled By: hx235 fbshipit-source-id: e5b4ee7001af26f2ee0377bc6334f43b2a527388
2023-05-25 17:16:58 +00:00
file_read_hist, skip_filters, level,
true /* prefetch_index_and_filter_in_cache */,
Multi file concurrency in MultiGet using coroutines and async IO (#9968) Summary: This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code. A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest. TODO: 1. Figure out how to build it in CircleCI (requires some dependencies to be installed) 2. Do some stress testing with coroutines enabled No regression in synchronous MultiGet between this branch and main - ``` ./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics ``` Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)``` Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)``` More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file. 1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) - No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)``` Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)``` 2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file - No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)``` Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)``` 3. Single thread CPU bound workload with ~2 key overlap/file - No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)``` Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)``` 4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file - No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ``` Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968 Reviewed By: akankshamahajan15 Differential Revision: D36348563 Pulled By: anand1976 fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
2022-05-19 22:36:27 +00:00
0 /*max_file_size_for_l0_meta_pin*/, file_meta.temperature);
TEST_SYNC_POINT_CALLBACK("TableCache::MultiGet:FindTable", &s);
if (s.ok()) {
Major Cache refactoring, CPU efficiency improvement (#10975) Summary: This is several refactorings bundled into one to avoid having to incrementally re-modify uses of Cache several times. Overall, there are breaking changes to Cache class, and it becomes more of low-level interface for implementing caches, especially block cache. New internal APIs make using Cache cleaner than before, and more insulated from block cache evolution. Hopefully, this is the last really big block cache refactoring, because of rather effectively decoupling the implementations from the uses. This change also removes the EXPERIMENTAL designation on the SecondaryCache support in Cache. It seems reasonably mature at this point but still subject to change/evolution (as I warn in the API docs for Cache). The high-level motivation for this refactoring is to minimize code duplication / compounding complexity in adding SecondaryCache support to HyperClockCache (in a later PR). Other benefits listed below. * static_cast lines of code +29 -35 (net removed 6) * reinterpret_cast lines of code +6 -32 (net removed 26) ## cache.h and secondary_cache.h * Always use CacheItemHelper with entries instead of just a Deleter. There are several motivations / justifications: * Simpler for implementations to deal with just one Insert and one Lookup. * Simpler and more efficient implementation because we don't have to track which entries are using helpers and which are using deleters * Gets rid of hack to classify cache entries by their deleter. Instead, the CacheItemHelper includes a CacheEntryRole. This simplifies a lot of code (cache_entry_roles.h almost eliminated). Fixes https://github.com/facebook/rocksdb/issues/9428. * Makes it trivial to adjust SecondaryCache behavior based on kind of block (e.g. don't re-compress filter blocks). * It is arguably less convenient for many direct users of Cache, but direct users of Cache are now rare with introduction of typed_cache.h (below). * I considered and rejected an alternative approach in which we reduce customizability by assuming each secondary cache compatible value starts with a Slice referencing the uncompressed block contents (already true or mostly true), but we apparently intend to stack secondary caches. Saving an entry from a compressed secondary to a lower tier requires custom handling offered by SaveToCallback, etc. * Make CreateCallback part of the helper and introduce CreateContext to work with it (alternative to https://github.com/facebook/rocksdb/issues/10562). This cleans up the interface while still allowing context to be provided for loading/parsing values into primary cache. This model works for async lookup in BlockBasedTable reader (reader owns a CreateContext) under the assumption that it always waits on secondary cache operations to finish. (Otherwise, the CreateContext could be destroyed while async operation depending on it continues.) This likely contributes most to the observed performance improvement because it saves an std::function backed by a heap allocation. * Use char* for serialized data, e.g. in SaveToCallback, where void* was confusingly used. (We use `char*` for serialized byte data all over RocksDB, with many advantages over `void*`. `memcpy` etc. are legacy APIs that should not be mimicked.) * Add a type alias Cache::ObjectPtr = void*, so that we can better indicate the intent of the void* when it is to be the object associated with a Cache entry. Related: started (but did not complete) a refactoring to move away from "value" of a cache entry toward "object" or "obj". (It is confusing to call Cache a key-value store (like DB) when it is really storing arbitrary in-memory objects, not byte strings.) * Remove unnecessary key param from DeleterFn. This is good for efficiency in HyperClockCache, which does not directly store the cache key in memory. (Alternative to https://github.com/facebook/rocksdb/issues/10774) * Add allocator to Cache DeleterFn. This is a kind of future-proofing change in case we get more serious about using the Cache allocator for memory tracked by the Cache. Right now, only the uncompressed block contents are allocated using the allocator, and a pointer to that allocator is saved as part of the cached object so that the deleter can use it. (See CacheAllocationPtr.) If in the future we are able to "flatten out" our Cache objects some more, it would be good not to have to track the allocator as part of each object. * Removes legacy `ApplyToAllCacheEntries` and changes `ApplyToAllEntries` signature for Deleter->CacheItemHelper change. ## typed_cache.h Adds various "typed" interfaces to the Cache as internal APIs, so that most uses of Cache can use simple type safe code without casting and without explicit deleters, etc. Almost all of the non-test, non-glue code uses of Cache have been migrated. (Follow-up work: CompressedSecondaryCache deserves deeper attention to migrate.) This change expands RocksDB's internal usage of metaprogramming and SFINAE (https://en.cppreference.com/w/cpp/language/sfinae). The existing usages of Cache are divided up at a high level into these new interfaces. See updated existing uses of Cache for examples of how these are used. * PlaceholderCacheInterface - Used for making cache reservations, with entries that have a charge but no value. * BasicTypedCacheInterface<TValue> - Used for primary cache storage of objects of type TValue, which can be cleaned up with std::default_delete<TValue>. The role is provided by TValue::kCacheEntryRole or given in an optional template parameter. * FullTypedCacheInterface<TValue, TCreateContext> - Used for secondary cache compatible storage of objects of type TValue. In addition to BasicTypedCacheInterface constraints, we require TValue::ContentSlice() to return persistable data. This simplifies usage for the normal case of simple secondary cache compatibility (can give you a Slice to the data already in memory). In addition to TCreateContext performing the role of Cache::CreateContext, it is also expected to provide a factory function for creating TValue. * For each of these, there's a "Shared" version (e.g. FullTypedSharedCacheInterface) that holds a shared_ptr to the Cache, rather than assuming external ownership by holding only a raw `Cache*`. These interfaces introduce specific handle types for each interface instantiation, so that it's easy to see what kind of object is controlled by a handle. (Ultimately, this might not be worth the extra complexity, but it seems OK so far.) Note: I attempted to make the cache 'charge' automatically inferred from the cache object type, such as by expecting an ApproximateMemoryUsage() function, but this is not so clean because there are cases where we need to compute the charge ahead of time and don't want to re-compute it. ## block_cache.h This header is essentially the replacement for the old block_like_traits.h. It includes various things to support block cache access with typed_cache.h for block-based table. ## block_based_table_reader.cc Before this change, accessing the block cache here was an awkward mix of static polymorphism (template TBlocklike) and switch-case on a dynamic BlockType value. This change mostly unifies on static polymorphism, relying on minor hacks in block_cache.h to distinguish variants of Block. We still check BlockType in some places (especially for stats, which could be improved in follow-up work) but at least the BlockType is a static constant from the template parameter. (No more awkward partial redundancy between static and dynamic info.) This likely contributes to the overall performance improvement, but hasn't been tested in isolation. The other key source of simplification here is a more unified system of creating block cache objects: for directly populating from primary cache and for promotion from secondary cache. Both use BlockCreateContext, for context and for factory functions. ## block_based_table_builder.cc, cache_dump_load_impl.cc Before this change, warming caches was super ugly code. Both of these source files had switch statements to basically transition from the dynamic BlockType world to the static TBlocklike world. None of that mess is needed anymore as there's a new, untyped WarmInCache function that handles all the details just as promotion from SecondaryCache would. (Fixes `TODO akanksha: Dedup below code` in block_based_table_builder.cc.) ## Everything else Mostly just updating Cache users to use new typed APIs when reasonably possible, or changed Cache APIs when not. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10975 Test Plan: tests updated Performance test setup similar to https://github.com/facebook/rocksdb/issues/10626 (by cache size, LRUCache when not "hyper" for HyperClockCache): 34MB 1thread base.hyper -> kops/s: 0.745 io_bytes/op: 2.52504e+06 miss_ratio: 0.140906 max_rss_mb: 76.4844 34MB 1thread new.hyper -> kops/s: 0.751 io_bytes/op: 2.5123e+06 miss_ratio: 0.140161 max_rss_mb: 79.3594 34MB 1thread base -> kops/s: 0.254 io_bytes/op: 1.36073e+07 miss_ratio: 0.918818 max_rss_mb: 45.9297 34MB 1thread new -> kops/s: 0.252 io_bytes/op: 1.36157e+07 miss_ratio: 0.918999 max_rss_mb: 44.1523 34MB 32thread base.hyper -> kops/s: 7.272 io_bytes/op: 2.88323e+06 miss_ratio: 0.162532 max_rss_mb: 516.602 34MB 32thread new.hyper -> kops/s: 7.214 io_bytes/op: 2.99046e+06 miss_ratio: 0.168818 max_rss_mb: 518.293 34MB 32thread base -> kops/s: 3.528 io_bytes/op: 1.35722e+07 miss_ratio: 0.914691 max_rss_mb: 264.926 34MB 32thread new -> kops/s: 3.604 io_bytes/op: 1.35744e+07 miss_ratio: 0.915054 max_rss_mb: 264.488 233MB 1thread base.hyper -> kops/s: 53.909 io_bytes/op: 2552.35 miss_ratio: 0.0440566 max_rss_mb: 241.984 233MB 1thread new.hyper -> kops/s: 62.792 io_bytes/op: 2549.79 miss_ratio: 0.044043 max_rss_mb: 241.922 233MB 1thread base -> kops/s: 1.197 io_bytes/op: 2.75173e+06 miss_ratio: 0.103093 max_rss_mb: 241.559 233MB 1thread new -> kops/s: 1.199 io_bytes/op: 2.73723e+06 miss_ratio: 0.10305 max_rss_mb: 240.93 233MB 32thread base.hyper -> kops/s: 1298.69 io_bytes/op: 2539.12 miss_ratio: 0.0440307 max_rss_mb: 371.418 233MB 32thread new.hyper -> kops/s: 1421.35 io_bytes/op: 2538.75 miss_ratio: 0.0440307 max_rss_mb: 347.273 233MB 32thread base -> kops/s: 9.693 io_bytes/op: 2.77304e+06 miss_ratio: 0.103745 max_rss_mb: 569.691 233MB 32thread new -> kops/s: 9.75 io_bytes/op: 2.77559e+06 miss_ratio: 0.103798 max_rss_mb: 552.82 1597MB 1thread base.hyper -> kops/s: 58.607 io_bytes/op: 1449.14 miss_ratio: 0.0249324 max_rss_mb: 1583.55 1597MB 1thread new.hyper -> kops/s: 69.6 io_bytes/op: 1434.89 miss_ratio: 0.0247167 max_rss_mb: 1584.02 1597MB 1thread base -> kops/s: 60.478 io_bytes/op: 1421.28 miss_ratio: 0.024452 max_rss_mb: 1589.45 1597MB 1thread new -> kops/s: 63.973 io_bytes/op: 1416.07 miss_ratio: 0.0243766 max_rss_mb: 1589.24 1597MB 32thread base.hyper -> kops/s: 1436.2 io_bytes/op: 1357.93 miss_ratio: 0.0235353 max_rss_mb: 1692.92 1597MB 32thread new.hyper -> kops/s: 1605.03 io_bytes/op: 1358.04 miss_ratio: 0.023538 max_rss_mb: 1702.78 1597MB 32thread base -> kops/s: 280.059 io_bytes/op: 1350.34 miss_ratio: 0.023289 max_rss_mb: 1675.36 1597MB 32thread new -> kops/s: 283.125 io_bytes/op: 1351.05 miss_ratio: 0.0232797 max_rss_mb: 1703.83 Almost uniformly improving over base revision, especially for hot paths with HyperClockCache, up to 12% higher throughput seen (1597MB, 32thread, hyper). The improvement for that is likely coming from much simplified code for providing context for secondary cache promotion (CreateCallback/CreateContext), and possibly from less branching in block_based_table_reader. And likely a small improvement from not reconstituting key for DeleterFn. Reviewed By: anand1976 Differential Revision: D42417818 Pulled By: pdillinger fbshipit-source-id: f86bfdd584dce27c028b151ba56818ad14f7a432
2023-01-11 22:20:40 +00:00
t = cache_.Value(handle);
Multi file concurrency in MultiGet using coroutines and async IO (#9968) Summary: This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code. A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest. TODO: 1. Figure out how to build it in CircleCI (requires some dependencies to be installed) 2. Do some stress testing with coroutines enabled No regression in synchronous MultiGet between this branch and main - ``` ./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics ``` Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)``` Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)``` More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file. 1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) - No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)``` Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)``` 2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file - No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)``` Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)``` 3. Single thread CPU bound workload with ~2 key overlap/file - No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)``` Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)``` 4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file - No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ``` Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968 Reviewed By: akankshamahajan15 Differential Revision: D36348563 Pulled By: anand1976 fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
2022-05-19 22:36:27 +00:00
assert(t);
}
}
if (s.ok() && !options.ignore_range_deletions && !skip_range_deletions) {
UpdateRangeTombstoneSeqnums(options, t, table_range);
Multi file concurrency in MultiGet using coroutines and async IO (#9968) Summary: This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code. A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest. TODO: 1. Figure out how to build it in CircleCI (requires some dependencies to be installed) 2. Do some stress testing with coroutines enabled No regression in synchronous MultiGet between this branch and main - ``` ./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics ``` Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)``` Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)``` More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file. 1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) - No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)``` Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)``` 2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file - No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)``` Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)``` 3. Single thread CPU bound workload with ~2 key overlap/file - No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)``` Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)``` 4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file - No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ``` Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968 Reviewed By: akankshamahajan15 Differential Revision: D36348563 Pulled By: anand1976 fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
2022-05-19 22:36:27 +00:00
}
if (s.ok()) {
CO_AWAIT(t->MultiGet)
(options, &table_range, prefix_extractor.get(), skip_filters);
} else if (options.read_tier == kBlockCacheTier && s.IsIncomplete()) {
for (auto iter = table_range.begin(); iter != table_range.end(); ++iter) {
Status* status = iter->s;
if (status->IsIncomplete()) {
// Couldn't find Table in cache but treat as kFound if no_io set
iter->get_context->MarkKeyMayExist();
s = Status::OK();
}
}
}
}
if (lookup_row_cache) {
size_t row_idx = 0;
Major Cache refactoring, CPU efficiency improvement (#10975) Summary: This is several refactorings bundled into one to avoid having to incrementally re-modify uses of Cache several times. Overall, there are breaking changes to Cache class, and it becomes more of low-level interface for implementing caches, especially block cache. New internal APIs make using Cache cleaner than before, and more insulated from block cache evolution. Hopefully, this is the last really big block cache refactoring, because of rather effectively decoupling the implementations from the uses. This change also removes the EXPERIMENTAL designation on the SecondaryCache support in Cache. It seems reasonably mature at this point but still subject to change/evolution (as I warn in the API docs for Cache). The high-level motivation for this refactoring is to minimize code duplication / compounding complexity in adding SecondaryCache support to HyperClockCache (in a later PR). Other benefits listed below. * static_cast lines of code +29 -35 (net removed 6) * reinterpret_cast lines of code +6 -32 (net removed 26) ## cache.h and secondary_cache.h * Always use CacheItemHelper with entries instead of just a Deleter. There are several motivations / justifications: * Simpler for implementations to deal with just one Insert and one Lookup. * Simpler and more efficient implementation because we don't have to track which entries are using helpers and which are using deleters * Gets rid of hack to classify cache entries by their deleter. Instead, the CacheItemHelper includes a CacheEntryRole. This simplifies a lot of code (cache_entry_roles.h almost eliminated). Fixes https://github.com/facebook/rocksdb/issues/9428. * Makes it trivial to adjust SecondaryCache behavior based on kind of block (e.g. don't re-compress filter blocks). * It is arguably less convenient for many direct users of Cache, but direct users of Cache are now rare with introduction of typed_cache.h (below). * I considered and rejected an alternative approach in which we reduce customizability by assuming each secondary cache compatible value starts with a Slice referencing the uncompressed block contents (already true or mostly true), but we apparently intend to stack secondary caches. Saving an entry from a compressed secondary to a lower tier requires custom handling offered by SaveToCallback, etc. * Make CreateCallback part of the helper and introduce CreateContext to work with it (alternative to https://github.com/facebook/rocksdb/issues/10562). This cleans up the interface while still allowing context to be provided for loading/parsing values into primary cache. This model works for async lookup in BlockBasedTable reader (reader owns a CreateContext) under the assumption that it always waits on secondary cache operations to finish. (Otherwise, the CreateContext could be destroyed while async operation depending on it continues.) This likely contributes most to the observed performance improvement because it saves an std::function backed by a heap allocation. * Use char* for serialized data, e.g. in SaveToCallback, where void* was confusingly used. (We use `char*` for serialized byte data all over RocksDB, with many advantages over `void*`. `memcpy` etc. are legacy APIs that should not be mimicked.) * Add a type alias Cache::ObjectPtr = void*, so that we can better indicate the intent of the void* when it is to be the object associated with a Cache entry. Related: started (but did not complete) a refactoring to move away from "value" of a cache entry toward "object" or "obj". (It is confusing to call Cache a key-value store (like DB) when it is really storing arbitrary in-memory objects, not byte strings.) * Remove unnecessary key param from DeleterFn. This is good for efficiency in HyperClockCache, which does not directly store the cache key in memory. (Alternative to https://github.com/facebook/rocksdb/issues/10774) * Add allocator to Cache DeleterFn. This is a kind of future-proofing change in case we get more serious about using the Cache allocator for memory tracked by the Cache. Right now, only the uncompressed block contents are allocated using the allocator, and a pointer to that allocator is saved as part of the cached object so that the deleter can use it. (See CacheAllocationPtr.) If in the future we are able to "flatten out" our Cache objects some more, it would be good not to have to track the allocator as part of each object. * Removes legacy `ApplyToAllCacheEntries` and changes `ApplyToAllEntries` signature for Deleter->CacheItemHelper change. ## typed_cache.h Adds various "typed" interfaces to the Cache as internal APIs, so that most uses of Cache can use simple type safe code without casting and without explicit deleters, etc. Almost all of the non-test, non-glue code uses of Cache have been migrated. (Follow-up work: CompressedSecondaryCache deserves deeper attention to migrate.) This change expands RocksDB's internal usage of metaprogramming and SFINAE (https://en.cppreference.com/w/cpp/language/sfinae). The existing usages of Cache are divided up at a high level into these new interfaces. See updated existing uses of Cache for examples of how these are used. * PlaceholderCacheInterface - Used for making cache reservations, with entries that have a charge but no value. * BasicTypedCacheInterface<TValue> - Used for primary cache storage of objects of type TValue, which can be cleaned up with std::default_delete<TValue>. The role is provided by TValue::kCacheEntryRole or given in an optional template parameter. * FullTypedCacheInterface<TValue, TCreateContext> - Used for secondary cache compatible storage of objects of type TValue. In addition to BasicTypedCacheInterface constraints, we require TValue::ContentSlice() to return persistable data. This simplifies usage for the normal case of simple secondary cache compatibility (can give you a Slice to the data already in memory). In addition to TCreateContext performing the role of Cache::CreateContext, it is also expected to provide a factory function for creating TValue. * For each of these, there's a "Shared" version (e.g. FullTypedSharedCacheInterface) that holds a shared_ptr to the Cache, rather than assuming external ownership by holding only a raw `Cache*`. These interfaces introduce specific handle types for each interface instantiation, so that it's easy to see what kind of object is controlled by a handle. (Ultimately, this might not be worth the extra complexity, but it seems OK so far.) Note: I attempted to make the cache 'charge' automatically inferred from the cache object type, such as by expecting an ApproximateMemoryUsage() function, but this is not so clean because there are cases where we need to compute the charge ahead of time and don't want to re-compute it. ## block_cache.h This header is essentially the replacement for the old block_like_traits.h. It includes various things to support block cache access with typed_cache.h for block-based table. ## block_based_table_reader.cc Before this change, accessing the block cache here was an awkward mix of static polymorphism (template TBlocklike) and switch-case on a dynamic BlockType value. This change mostly unifies on static polymorphism, relying on minor hacks in block_cache.h to distinguish variants of Block. We still check BlockType in some places (especially for stats, which could be improved in follow-up work) but at least the BlockType is a static constant from the template parameter. (No more awkward partial redundancy between static and dynamic info.) This likely contributes to the overall performance improvement, but hasn't been tested in isolation. The other key source of simplification here is a more unified system of creating block cache objects: for directly populating from primary cache and for promotion from secondary cache. Both use BlockCreateContext, for context and for factory functions. ## block_based_table_builder.cc, cache_dump_load_impl.cc Before this change, warming caches was super ugly code. Both of these source files had switch statements to basically transition from the dynamic BlockType world to the static TBlocklike world. None of that mess is needed anymore as there's a new, untyped WarmInCache function that handles all the details just as promotion from SecondaryCache would. (Fixes `TODO akanksha: Dedup below code` in block_based_table_builder.cc.) ## Everything else Mostly just updating Cache users to use new typed APIs when reasonably possible, or changed Cache APIs when not. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10975 Test Plan: tests updated Performance test setup similar to https://github.com/facebook/rocksdb/issues/10626 (by cache size, LRUCache when not "hyper" for HyperClockCache): 34MB 1thread base.hyper -> kops/s: 0.745 io_bytes/op: 2.52504e+06 miss_ratio: 0.140906 max_rss_mb: 76.4844 34MB 1thread new.hyper -> kops/s: 0.751 io_bytes/op: 2.5123e+06 miss_ratio: 0.140161 max_rss_mb: 79.3594 34MB 1thread base -> kops/s: 0.254 io_bytes/op: 1.36073e+07 miss_ratio: 0.918818 max_rss_mb: 45.9297 34MB 1thread new -> kops/s: 0.252 io_bytes/op: 1.36157e+07 miss_ratio: 0.918999 max_rss_mb: 44.1523 34MB 32thread base.hyper -> kops/s: 7.272 io_bytes/op: 2.88323e+06 miss_ratio: 0.162532 max_rss_mb: 516.602 34MB 32thread new.hyper -> kops/s: 7.214 io_bytes/op: 2.99046e+06 miss_ratio: 0.168818 max_rss_mb: 518.293 34MB 32thread base -> kops/s: 3.528 io_bytes/op: 1.35722e+07 miss_ratio: 0.914691 max_rss_mb: 264.926 34MB 32thread new -> kops/s: 3.604 io_bytes/op: 1.35744e+07 miss_ratio: 0.915054 max_rss_mb: 264.488 233MB 1thread base.hyper -> kops/s: 53.909 io_bytes/op: 2552.35 miss_ratio: 0.0440566 max_rss_mb: 241.984 233MB 1thread new.hyper -> kops/s: 62.792 io_bytes/op: 2549.79 miss_ratio: 0.044043 max_rss_mb: 241.922 233MB 1thread base -> kops/s: 1.197 io_bytes/op: 2.75173e+06 miss_ratio: 0.103093 max_rss_mb: 241.559 233MB 1thread new -> kops/s: 1.199 io_bytes/op: 2.73723e+06 miss_ratio: 0.10305 max_rss_mb: 240.93 233MB 32thread base.hyper -> kops/s: 1298.69 io_bytes/op: 2539.12 miss_ratio: 0.0440307 max_rss_mb: 371.418 233MB 32thread new.hyper -> kops/s: 1421.35 io_bytes/op: 2538.75 miss_ratio: 0.0440307 max_rss_mb: 347.273 233MB 32thread base -> kops/s: 9.693 io_bytes/op: 2.77304e+06 miss_ratio: 0.103745 max_rss_mb: 569.691 233MB 32thread new -> kops/s: 9.75 io_bytes/op: 2.77559e+06 miss_ratio: 0.103798 max_rss_mb: 552.82 1597MB 1thread base.hyper -> kops/s: 58.607 io_bytes/op: 1449.14 miss_ratio: 0.0249324 max_rss_mb: 1583.55 1597MB 1thread new.hyper -> kops/s: 69.6 io_bytes/op: 1434.89 miss_ratio: 0.0247167 max_rss_mb: 1584.02 1597MB 1thread base -> kops/s: 60.478 io_bytes/op: 1421.28 miss_ratio: 0.024452 max_rss_mb: 1589.45 1597MB 1thread new -> kops/s: 63.973 io_bytes/op: 1416.07 miss_ratio: 0.0243766 max_rss_mb: 1589.24 1597MB 32thread base.hyper -> kops/s: 1436.2 io_bytes/op: 1357.93 miss_ratio: 0.0235353 max_rss_mb: 1692.92 1597MB 32thread new.hyper -> kops/s: 1605.03 io_bytes/op: 1358.04 miss_ratio: 0.023538 max_rss_mb: 1702.78 1597MB 32thread base -> kops/s: 280.059 io_bytes/op: 1350.34 miss_ratio: 0.023289 max_rss_mb: 1675.36 1597MB 32thread new -> kops/s: 283.125 io_bytes/op: 1351.05 miss_ratio: 0.0232797 max_rss_mb: 1703.83 Almost uniformly improving over base revision, especially for hot paths with HyperClockCache, up to 12% higher throughput seen (1597MB, 32thread, hyper). The improvement for that is likely coming from much simplified code for providing context for secondary cache promotion (CreateCallback/CreateContext), and possibly from less branching in block_based_table_reader. And likely a small improvement from not reconstituting key for DeleterFn. Reviewed By: anand1976 Differential Revision: D42417818 Pulled By: pdillinger fbshipit-source-id: f86bfdd584dce27c028b151ba56818ad14f7a432
2023-01-11 22:20:40 +00:00
RowCacheInterface row_cache{ioptions_.row_cache.get()};
Multi file concurrency in MultiGet using coroutines and async IO (#9968) Summary: This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code. A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest. TODO: 1. Figure out how to build it in CircleCI (requires some dependencies to be installed) 2. Do some stress testing with coroutines enabled No regression in synchronous MultiGet between this branch and main - ``` ./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics ``` Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)``` Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)``` More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file. 1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) - No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)``` Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)``` 2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file - No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)``` Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)``` 3. Single thread CPU bound workload with ~2 key overlap/file - No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)``` Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)``` 4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file - No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ``` Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968 Reviewed By: akankshamahajan15 Differential Revision: D36348563 Pulled By: anand1976 fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
2022-05-19 22:36:27 +00:00
for (auto miter = table_range.begin(); miter != table_range.end();
++miter) {
std::string& row_cache_entry = row_cache_entries[row_idx++];
const Slice& user_key = miter->ukey_with_ts;
GetContext* get_context = miter->get_context;
get_context->SetReplayLog(nullptr);
// Compute row cache key.
row_cache_key.TrimAppend(row_cache_key_prefix_size, user_key.data(),
user_key.size());
// Put the replay log in row cache only if something was found.
if (s.ok() && !row_cache_entry.empty()) {
size_t charge = row_cache_entry.capacity() + sizeof(std::string);
Major Cache refactoring, CPU efficiency improvement (#10975) Summary: This is several refactorings bundled into one to avoid having to incrementally re-modify uses of Cache several times. Overall, there are breaking changes to Cache class, and it becomes more of low-level interface for implementing caches, especially block cache. New internal APIs make using Cache cleaner than before, and more insulated from block cache evolution. Hopefully, this is the last really big block cache refactoring, because of rather effectively decoupling the implementations from the uses. This change also removes the EXPERIMENTAL designation on the SecondaryCache support in Cache. It seems reasonably mature at this point but still subject to change/evolution (as I warn in the API docs for Cache). The high-level motivation for this refactoring is to minimize code duplication / compounding complexity in adding SecondaryCache support to HyperClockCache (in a later PR). Other benefits listed below. * static_cast lines of code +29 -35 (net removed 6) * reinterpret_cast lines of code +6 -32 (net removed 26) ## cache.h and secondary_cache.h * Always use CacheItemHelper with entries instead of just a Deleter. There are several motivations / justifications: * Simpler for implementations to deal with just one Insert and one Lookup. * Simpler and more efficient implementation because we don't have to track which entries are using helpers and which are using deleters * Gets rid of hack to classify cache entries by their deleter. Instead, the CacheItemHelper includes a CacheEntryRole. This simplifies a lot of code (cache_entry_roles.h almost eliminated). Fixes https://github.com/facebook/rocksdb/issues/9428. * Makes it trivial to adjust SecondaryCache behavior based on kind of block (e.g. don't re-compress filter blocks). * It is arguably less convenient for many direct users of Cache, but direct users of Cache are now rare with introduction of typed_cache.h (below). * I considered and rejected an alternative approach in which we reduce customizability by assuming each secondary cache compatible value starts with a Slice referencing the uncompressed block contents (already true or mostly true), but we apparently intend to stack secondary caches. Saving an entry from a compressed secondary to a lower tier requires custom handling offered by SaveToCallback, etc. * Make CreateCallback part of the helper and introduce CreateContext to work with it (alternative to https://github.com/facebook/rocksdb/issues/10562). This cleans up the interface while still allowing context to be provided for loading/parsing values into primary cache. This model works for async lookup in BlockBasedTable reader (reader owns a CreateContext) under the assumption that it always waits on secondary cache operations to finish. (Otherwise, the CreateContext could be destroyed while async operation depending on it continues.) This likely contributes most to the observed performance improvement because it saves an std::function backed by a heap allocation. * Use char* for serialized data, e.g. in SaveToCallback, where void* was confusingly used. (We use `char*` for serialized byte data all over RocksDB, with many advantages over `void*`. `memcpy` etc. are legacy APIs that should not be mimicked.) * Add a type alias Cache::ObjectPtr = void*, so that we can better indicate the intent of the void* when it is to be the object associated with a Cache entry. Related: started (but did not complete) a refactoring to move away from "value" of a cache entry toward "object" or "obj". (It is confusing to call Cache a key-value store (like DB) when it is really storing arbitrary in-memory objects, not byte strings.) * Remove unnecessary key param from DeleterFn. This is good for efficiency in HyperClockCache, which does not directly store the cache key in memory. (Alternative to https://github.com/facebook/rocksdb/issues/10774) * Add allocator to Cache DeleterFn. This is a kind of future-proofing change in case we get more serious about using the Cache allocator for memory tracked by the Cache. Right now, only the uncompressed block contents are allocated using the allocator, and a pointer to that allocator is saved as part of the cached object so that the deleter can use it. (See CacheAllocationPtr.) If in the future we are able to "flatten out" our Cache objects some more, it would be good not to have to track the allocator as part of each object. * Removes legacy `ApplyToAllCacheEntries` and changes `ApplyToAllEntries` signature for Deleter->CacheItemHelper change. ## typed_cache.h Adds various "typed" interfaces to the Cache as internal APIs, so that most uses of Cache can use simple type safe code without casting and without explicit deleters, etc. Almost all of the non-test, non-glue code uses of Cache have been migrated. (Follow-up work: CompressedSecondaryCache deserves deeper attention to migrate.) This change expands RocksDB's internal usage of metaprogramming and SFINAE (https://en.cppreference.com/w/cpp/language/sfinae). The existing usages of Cache are divided up at a high level into these new interfaces. See updated existing uses of Cache for examples of how these are used. * PlaceholderCacheInterface - Used for making cache reservations, with entries that have a charge but no value. * BasicTypedCacheInterface<TValue> - Used for primary cache storage of objects of type TValue, which can be cleaned up with std::default_delete<TValue>. The role is provided by TValue::kCacheEntryRole or given in an optional template parameter. * FullTypedCacheInterface<TValue, TCreateContext> - Used for secondary cache compatible storage of objects of type TValue. In addition to BasicTypedCacheInterface constraints, we require TValue::ContentSlice() to return persistable data. This simplifies usage for the normal case of simple secondary cache compatibility (can give you a Slice to the data already in memory). In addition to TCreateContext performing the role of Cache::CreateContext, it is also expected to provide a factory function for creating TValue. * For each of these, there's a "Shared" version (e.g. FullTypedSharedCacheInterface) that holds a shared_ptr to the Cache, rather than assuming external ownership by holding only a raw `Cache*`. These interfaces introduce specific handle types for each interface instantiation, so that it's easy to see what kind of object is controlled by a handle. (Ultimately, this might not be worth the extra complexity, but it seems OK so far.) Note: I attempted to make the cache 'charge' automatically inferred from the cache object type, such as by expecting an ApproximateMemoryUsage() function, but this is not so clean because there are cases where we need to compute the charge ahead of time and don't want to re-compute it. ## block_cache.h This header is essentially the replacement for the old block_like_traits.h. It includes various things to support block cache access with typed_cache.h for block-based table. ## block_based_table_reader.cc Before this change, accessing the block cache here was an awkward mix of static polymorphism (template TBlocklike) and switch-case on a dynamic BlockType value. This change mostly unifies on static polymorphism, relying on minor hacks in block_cache.h to distinguish variants of Block. We still check BlockType in some places (especially for stats, which could be improved in follow-up work) but at least the BlockType is a static constant from the template parameter. (No more awkward partial redundancy between static and dynamic info.) This likely contributes to the overall performance improvement, but hasn't been tested in isolation. The other key source of simplification here is a more unified system of creating block cache objects: for directly populating from primary cache and for promotion from secondary cache. Both use BlockCreateContext, for context and for factory functions. ## block_based_table_builder.cc, cache_dump_load_impl.cc Before this change, warming caches was super ugly code. Both of these source files had switch statements to basically transition from the dynamic BlockType world to the static TBlocklike world. None of that mess is needed anymore as there's a new, untyped WarmInCache function that handles all the details just as promotion from SecondaryCache would. (Fixes `TODO akanksha: Dedup below code` in block_based_table_builder.cc.) ## Everything else Mostly just updating Cache users to use new typed APIs when reasonably possible, or changed Cache APIs when not. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10975 Test Plan: tests updated Performance test setup similar to https://github.com/facebook/rocksdb/issues/10626 (by cache size, LRUCache when not "hyper" for HyperClockCache): 34MB 1thread base.hyper -> kops/s: 0.745 io_bytes/op: 2.52504e+06 miss_ratio: 0.140906 max_rss_mb: 76.4844 34MB 1thread new.hyper -> kops/s: 0.751 io_bytes/op: 2.5123e+06 miss_ratio: 0.140161 max_rss_mb: 79.3594 34MB 1thread base -> kops/s: 0.254 io_bytes/op: 1.36073e+07 miss_ratio: 0.918818 max_rss_mb: 45.9297 34MB 1thread new -> kops/s: 0.252 io_bytes/op: 1.36157e+07 miss_ratio: 0.918999 max_rss_mb: 44.1523 34MB 32thread base.hyper -> kops/s: 7.272 io_bytes/op: 2.88323e+06 miss_ratio: 0.162532 max_rss_mb: 516.602 34MB 32thread new.hyper -> kops/s: 7.214 io_bytes/op: 2.99046e+06 miss_ratio: 0.168818 max_rss_mb: 518.293 34MB 32thread base -> kops/s: 3.528 io_bytes/op: 1.35722e+07 miss_ratio: 0.914691 max_rss_mb: 264.926 34MB 32thread new -> kops/s: 3.604 io_bytes/op: 1.35744e+07 miss_ratio: 0.915054 max_rss_mb: 264.488 233MB 1thread base.hyper -> kops/s: 53.909 io_bytes/op: 2552.35 miss_ratio: 0.0440566 max_rss_mb: 241.984 233MB 1thread new.hyper -> kops/s: 62.792 io_bytes/op: 2549.79 miss_ratio: 0.044043 max_rss_mb: 241.922 233MB 1thread base -> kops/s: 1.197 io_bytes/op: 2.75173e+06 miss_ratio: 0.103093 max_rss_mb: 241.559 233MB 1thread new -> kops/s: 1.199 io_bytes/op: 2.73723e+06 miss_ratio: 0.10305 max_rss_mb: 240.93 233MB 32thread base.hyper -> kops/s: 1298.69 io_bytes/op: 2539.12 miss_ratio: 0.0440307 max_rss_mb: 371.418 233MB 32thread new.hyper -> kops/s: 1421.35 io_bytes/op: 2538.75 miss_ratio: 0.0440307 max_rss_mb: 347.273 233MB 32thread base -> kops/s: 9.693 io_bytes/op: 2.77304e+06 miss_ratio: 0.103745 max_rss_mb: 569.691 233MB 32thread new -> kops/s: 9.75 io_bytes/op: 2.77559e+06 miss_ratio: 0.103798 max_rss_mb: 552.82 1597MB 1thread base.hyper -> kops/s: 58.607 io_bytes/op: 1449.14 miss_ratio: 0.0249324 max_rss_mb: 1583.55 1597MB 1thread new.hyper -> kops/s: 69.6 io_bytes/op: 1434.89 miss_ratio: 0.0247167 max_rss_mb: 1584.02 1597MB 1thread base -> kops/s: 60.478 io_bytes/op: 1421.28 miss_ratio: 0.024452 max_rss_mb: 1589.45 1597MB 1thread new -> kops/s: 63.973 io_bytes/op: 1416.07 miss_ratio: 0.0243766 max_rss_mb: 1589.24 1597MB 32thread base.hyper -> kops/s: 1436.2 io_bytes/op: 1357.93 miss_ratio: 0.0235353 max_rss_mb: 1692.92 1597MB 32thread new.hyper -> kops/s: 1605.03 io_bytes/op: 1358.04 miss_ratio: 0.023538 max_rss_mb: 1702.78 1597MB 32thread base -> kops/s: 280.059 io_bytes/op: 1350.34 miss_ratio: 0.023289 max_rss_mb: 1675.36 1597MB 32thread new -> kops/s: 283.125 io_bytes/op: 1351.05 miss_ratio: 0.0232797 max_rss_mb: 1703.83 Almost uniformly improving over base revision, especially for hot paths with HyperClockCache, up to 12% higher throughput seen (1597MB, 32thread, hyper). The improvement for that is likely coming from much simplified code for providing context for secondary cache promotion (CreateCallback/CreateContext), and possibly from less branching in block_based_table_reader. And likely a small improvement from not reconstituting key for DeleterFn. Reviewed By: anand1976 Differential Revision: D42417818 Pulled By: pdillinger fbshipit-source-id: f86bfdd584dce27c028b151ba56818ad14f7a432
2023-01-11 22:20:40 +00:00
auto row_ptr = new std::string(std::move(row_cache_entry));
Multi file concurrency in MultiGet using coroutines and async IO (#9968) Summary: This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code. A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest. TODO: 1. Figure out how to build it in CircleCI (requires some dependencies to be installed) 2. Do some stress testing with coroutines enabled No regression in synchronous MultiGet between this branch and main - ``` ./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics ``` Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)``` Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)``` More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file. 1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) - No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)``` Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)``` 2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file - No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)``` Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)``` 3. Single thread CPU bound workload with ~2 key overlap/file - No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)``` Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)``` 4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file - No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ``` Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968 Reviewed By: akankshamahajan15 Differential Revision: D36348563 Pulled By: anand1976 fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
2022-05-19 22:36:27 +00:00
// If row cache is full, it's OK.
Major Cache refactoring, CPU efficiency improvement (#10975) Summary: This is several refactorings bundled into one to avoid having to incrementally re-modify uses of Cache several times. Overall, there are breaking changes to Cache class, and it becomes more of low-level interface for implementing caches, especially block cache. New internal APIs make using Cache cleaner than before, and more insulated from block cache evolution. Hopefully, this is the last really big block cache refactoring, because of rather effectively decoupling the implementations from the uses. This change also removes the EXPERIMENTAL designation on the SecondaryCache support in Cache. It seems reasonably mature at this point but still subject to change/evolution (as I warn in the API docs for Cache). The high-level motivation for this refactoring is to minimize code duplication / compounding complexity in adding SecondaryCache support to HyperClockCache (in a later PR). Other benefits listed below. * static_cast lines of code +29 -35 (net removed 6) * reinterpret_cast lines of code +6 -32 (net removed 26) ## cache.h and secondary_cache.h * Always use CacheItemHelper with entries instead of just a Deleter. There are several motivations / justifications: * Simpler for implementations to deal with just one Insert and one Lookup. * Simpler and more efficient implementation because we don't have to track which entries are using helpers and which are using deleters * Gets rid of hack to classify cache entries by their deleter. Instead, the CacheItemHelper includes a CacheEntryRole. This simplifies a lot of code (cache_entry_roles.h almost eliminated). Fixes https://github.com/facebook/rocksdb/issues/9428. * Makes it trivial to adjust SecondaryCache behavior based on kind of block (e.g. don't re-compress filter blocks). * It is arguably less convenient for many direct users of Cache, but direct users of Cache are now rare with introduction of typed_cache.h (below). * I considered and rejected an alternative approach in which we reduce customizability by assuming each secondary cache compatible value starts with a Slice referencing the uncompressed block contents (already true or mostly true), but we apparently intend to stack secondary caches. Saving an entry from a compressed secondary to a lower tier requires custom handling offered by SaveToCallback, etc. * Make CreateCallback part of the helper and introduce CreateContext to work with it (alternative to https://github.com/facebook/rocksdb/issues/10562). This cleans up the interface while still allowing context to be provided for loading/parsing values into primary cache. This model works for async lookup in BlockBasedTable reader (reader owns a CreateContext) under the assumption that it always waits on secondary cache operations to finish. (Otherwise, the CreateContext could be destroyed while async operation depending on it continues.) This likely contributes most to the observed performance improvement because it saves an std::function backed by a heap allocation. * Use char* for serialized data, e.g. in SaveToCallback, where void* was confusingly used. (We use `char*` for serialized byte data all over RocksDB, with many advantages over `void*`. `memcpy` etc. are legacy APIs that should not be mimicked.) * Add a type alias Cache::ObjectPtr = void*, so that we can better indicate the intent of the void* when it is to be the object associated with a Cache entry. Related: started (but did not complete) a refactoring to move away from "value" of a cache entry toward "object" or "obj". (It is confusing to call Cache a key-value store (like DB) when it is really storing arbitrary in-memory objects, not byte strings.) * Remove unnecessary key param from DeleterFn. This is good for efficiency in HyperClockCache, which does not directly store the cache key in memory. (Alternative to https://github.com/facebook/rocksdb/issues/10774) * Add allocator to Cache DeleterFn. This is a kind of future-proofing change in case we get more serious about using the Cache allocator for memory tracked by the Cache. Right now, only the uncompressed block contents are allocated using the allocator, and a pointer to that allocator is saved as part of the cached object so that the deleter can use it. (See CacheAllocationPtr.) If in the future we are able to "flatten out" our Cache objects some more, it would be good not to have to track the allocator as part of each object. * Removes legacy `ApplyToAllCacheEntries` and changes `ApplyToAllEntries` signature for Deleter->CacheItemHelper change. ## typed_cache.h Adds various "typed" interfaces to the Cache as internal APIs, so that most uses of Cache can use simple type safe code without casting and without explicit deleters, etc. Almost all of the non-test, non-glue code uses of Cache have been migrated. (Follow-up work: CompressedSecondaryCache deserves deeper attention to migrate.) This change expands RocksDB's internal usage of metaprogramming and SFINAE (https://en.cppreference.com/w/cpp/language/sfinae). The existing usages of Cache are divided up at a high level into these new interfaces. See updated existing uses of Cache for examples of how these are used. * PlaceholderCacheInterface - Used for making cache reservations, with entries that have a charge but no value. * BasicTypedCacheInterface<TValue> - Used for primary cache storage of objects of type TValue, which can be cleaned up with std::default_delete<TValue>. The role is provided by TValue::kCacheEntryRole or given in an optional template parameter. * FullTypedCacheInterface<TValue, TCreateContext> - Used for secondary cache compatible storage of objects of type TValue. In addition to BasicTypedCacheInterface constraints, we require TValue::ContentSlice() to return persistable data. This simplifies usage for the normal case of simple secondary cache compatibility (can give you a Slice to the data already in memory). In addition to TCreateContext performing the role of Cache::CreateContext, it is also expected to provide a factory function for creating TValue. * For each of these, there's a "Shared" version (e.g. FullTypedSharedCacheInterface) that holds a shared_ptr to the Cache, rather than assuming external ownership by holding only a raw `Cache*`. These interfaces introduce specific handle types for each interface instantiation, so that it's easy to see what kind of object is controlled by a handle. (Ultimately, this might not be worth the extra complexity, but it seems OK so far.) Note: I attempted to make the cache 'charge' automatically inferred from the cache object type, such as by expecting an ApproximateMemoryUsage() function, but this is not so clean because there are cases where we need to compute the charge ahead of time and don't want to re-compute it. ## block_cache.h This header is essentially the replacement for the old block_like_traits.h. It includes various things to support block cache access with typed_cache.h for block-based table. ## block_based_table_reader.cc Before this change, accessing the block cache here was an awkward mix of static polymorphism (template TBlocklike) and switch-case on a dynamic BlockType value. This change mostly unifies on static polymorphism, relying on minor hacks in block_cache.h to distinguish variants of Block. We still check BlockType in some places (especially for stats, which could be improved in follow-up work) but at least the BlockType is a static constant from the template parameter. (No more awkward partial redundancy between static and dynamic info.) This likely contributes to the overall performance improvement, but hasn't been tested in isolation. The other key source of simplification here is a more unified system of creating block cache objects: for directly populating from primary cache and for promotion from secondary cache. Both use BlockCreateContext, for context and for factory functions. ## block_based_table_builder.cc, cache_dump_load_impl.cc Before this change, warming caches was super ugly code. Both of these source files had switch statements to basically transition from the dynamic BlockType world to the static TBlocklike world. None of that mess is needed anymore as there's a new, untyped WarmInCache function that handles all the details just as promotion from SecondaryCache would. (Fixes `TODO akanksha: Dedup below code` in block_based_table_builder.cc.) ## Everything else Mostly just updating Cache users to use new typed APIs when reasonably possible, or changed Cache APIs when not. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10975 Test Plan: tests updated Performance test setup similar to https://github.com/facebook/rocksdb/issues/10626 (by cache size, LRUCache when not "hyper" for HyperClockCache): 34MB 1thread base.hyper -> kops/s: 0.745 io_bytes/op: 2.52504e+06 miss_ratio: 0.140906 max_rss_mb: 76.4844 34MB 1thread new.hyper -> kops/s: 0.751 io_bytes/op: 2.5123e+06 miss_ratio: 0.140161 max_rss_mb: 79.3594 34MB 1thread base -> kops/s: 0.254 io_bytes/op: 1.36073e+07 miss_ratio: 0.918818 max_rss_mb: 45.9297 34MB 1thread new -> kops/s: 0.252 io_bytes/op: 1.36157e+07 miss_ratio: 0.918999 max_rss_mb: 44.1523 34MB 32thread base.hyper -> kops/s: 7.272 io_bytes/op: 2.88323e+06 miss_ratio: 0.162532 max_rss_mb: 516.602 34MB 32thread new.hyper -> kops/s: 7.214 io_bytes/op: 2.99046e+06 miss_ratio: 0.168818 max_rss_mb: 518.293 34MB 32thread base -> kops/s: 3.528 io_bytes/op: 1.35722e+07 miss_ratio: 0.914691 max_rss_mb: 264.926 34MB 32thread new -> kops/s: 3.604 io_bytes/op: 1.35744e+07 miss_ratio: 0.915054 max_rss_mb: 264.488 233MB 1thread base.hyper -> kops/s: 53.909 io_bytes/op: 2552.35 miss_ratio: 0.0440566 max_rss_mb: 241.984 233MB 1thread new.hyper -> kops/s: 62.792 io_bytes/op: 2549.79 miss_ratio: 0.044043 max_rss_mb: 241.922 233MB 1thread base -> kops/s: 1.197 io_bytes/op: 2.75173e+06 miss_ratio: 0.103093 max_rss_mb: 241.559 233MB 1thread new -> kops/s: 1.199 io_bytes/op: 2.73723e+06 miss_ratio: 0.10305 max_rss_mb: 240.93 233MB 32thread base.hyper -> kops/s: 1298.69 io_bytes/op: 2539.12 miss_ratio: 0.0440307 max_rss_mb: 371.418 233MB 32thread new.hyper -> kops/s: 1421.35 io_bytes/op: 2538.75 miss_ratio: 0.0440307 max_rss_mb: 347.273 233MB 32thread base -> kops/s: 9.693 io_bytes/op: 2.77304e+06 miss_ratio: 0.103745 max_rss_mb: 569.691 233MB 32thread new -> kops/s: 9.75 io_bytes/op: 2.77559e+06 miss_ratio: 0.103798 max_rss_mb: 552.82 1597MB 1thread base.hyper -> kops/s: 58.607 io_bytes/op: 1449.14 miss_ratio: 0.0249324 max_rss_mb: 1583.55 1597MB 1thread new.hyper -> kops/s: 69.6 io_bytes/op: 1434.89 miss_ratio: 0.0247167 max_rss_mb: 1584.02 1597MB 1thread base -> kops/s: 60.478 io_bytes/op: 1421.28 miss_ratio: 0.024452 max_rss_mb: 1589.45 1597MB 1thread new -> kops/s: 63.973 io_bytes/op: 1416.07 miss_ratio: 0.0243766 max_rss_mb: 1589.24 1597MB 32thread base.hyper -> kops/s: 1436.2 io_bytes/op: 1357.93 miss_ratio: 0.0235353 max_rss_mb: 1692.92 1597MB 32thread new.hyper -> kops/s: 1605.03 io_bytes/op: 1358.04 miss_ratio: 0.023538 max_rss_mb: 1702.78 1597MB 32thread base -> kops/s: 280.059 io_bytes/op: 1350.34 miss_ratio: 0.023289 max_rss_mb: 1675.36 1597MB 32thread new -> kops/s: 283.125 io_bytes/op: 1351.05 miss_ratio: 0.0232797 max_rss_mb: 1703.83 Almost uniformly improving over base revision, especially for hot paths with HyperClockCache, up to 12% higher throughput seen (1597MB, 32thread, hyper). The improvement for that is likely coming from much simplified code for providing context for secondary cache promotion (CreateCallback/CreateContext), and possibly from less branching in block_based_table_reader. And likely a small improvement from not reconstituting key for DeleterFn. Reviewed By: anand1976 Differential Revision: D42417818 Pulled By: pdillinger fbshipit-source-id: f86bfdd584dce27c028b151ba56818ad14f7a432
2023-01-11 22:20:40 +00:00
row_cache.Insert(row_cache_key.GetUserKey(), row_ptr, charge)
Multi file concurrency in MultiGet using coroutines and async IO (#9968) Summary: This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code. A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest. TODO: 1. Figure out how to build it in CircleCI (requires some dependencies to be installed) 2. Do some stress testing with coroutines enabled No regression in synchronous MultiGet between this branch and main - ``` ./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics ``` Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)``` Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)``` More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file. 1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) - No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)``` Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)``` 2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file - No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)``` Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)``` 3. Single thread CPU bound workload with ~2 key overlap/file - No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)``` Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)``` 4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file - No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ``` Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968 Reviewed By: akankshamahajan15 Differential Revision: D36348563 Pulled By: anand1976 fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
2022-05-19 22:36:27 +00:00
.PermitUncheckedError();
}
}
}
if (handle != nullptr) {
Major Cache refactoring, CPU efficiency improvement (#10975) Summary: This is several refactorings bundled into one to avoid having to incrementally re-modify uses of Cache several times. Overall, there are breaking changes to Cache class, and it becomes more of low-level interface for implementing caches, especially block cache. New internal APIs make using Cache cleaner than before, and more insulated from block cache evolution. Hopefully, this is the last really big block cache refactoring, because of rather effectively decoupling the implementations from the uses. This change also removes the EXPERIMENTAL designation on the SecondaryCache support in Cache. It seems reasonably mature at this point but still subject to change/evolution (as I warn in the API docs for Cache). The high-level motivation for this refactoring is to minimize code duplication / compounding complexity in adding SecondaryCache support to HyperClockCache (in a later PR). Other benefits listed below. * static_cast lines of code +29 -35 (net removed 6) * reinterpret_cast lines of code +6 -32 (net removed 26) ## cache.h and secondary_cache.h * Always use CacheItemHelper with entries instead of just a Deleter. There are several motivations / justifications: * Simpler for implementations to deal with just one Insert and one Lookup. * Simpler and more efficient implementation because we don't have to track which entries are using helpers and which are using deleters * Gets rid of hack to classify cache entries by their deleter. Instead, the CacheItemHelper includes a CacheEntryRole. This simplifies a lot of code (cache_entry_roles.h almost eliminated). Fixes https://github.com/facebook/rocksdb/issues/9428. * Makes it trivial to adjust SecondaryCache behavior based on kind of block (e.g. don't re-compress filter blocks). * It is arguably less convenient for many direct users of Cache, but direct users of Cache are now rare with introduction of typed_cache.h (below). * I considered and rejected an alternative approach in which we reduce customizability by assuming each secondary cache compatible value starts with a Slice referencing the uncompressed block contents (already true or mostly true), but we apparently intend to stack secondary caches. Saving an entry from a compressed secondary to a lower tier requires custom handling offered by SaveToCallback, etc. * Make CreateCallback part of the helper and introduce CreateContext to work with it (alternative to https://github.com/facebook/rocksdb/issues/10562). This cleans up the interface while still allowing context to be provided for loading/parsing values into primary cache. This model works for async lookup in BlockBasedTable reader (reader owns a CreateContext) under the assumption that it always waits on secondary cache operations to finish. (Otherwise, the CreateContext could be destroyed while async operation depending on it continues.) This likely contributes most to the observed performance improvement because it saves an std::function backed by a heap allocation. * Use char* for serialized data, e.g. in SaveToCallback, where void* was confusingly used. (We use `char*` for serialized byte data all over RocksDB, with many advantages over `void*`. `memcpy` etc. are legacy APIs that should not be mimicked.) * Add a type alias Cache::ObjectPtr = void*, so that we can better indicate the intent of the void* when it is to be the object associated with a Cache entry. Related: started (but did not complete) a refactoring to move away from "value" of a cache entry toward "object" or "obj". (It is confusing to call Cache a key-value store (like DB) when it is really storing arbitrary in-memory objects, not byte strings.) * Remove unnecessary key param from DeleterFn. This is good for efficiency in HyperClockCache, which does not directly store the cache key in memory. (Alternative to https://github.com/facebook/rocksdb/issues/10774) * Add allocator to Cache DeleterFn. This is a kind of future-proofing change in case we get more serious about using the Cache allocator for memory tracked by the Cache. Right now, only the uncompressed block contents are allocated using the allocator, and a pointer to that allocator is saved as part of the cached object so that the deleter can use it. (See CacheAllocationPtr.) If in the future we are able to "flatten out" our Cache objects some more, it would be good not to have to track the allocator as part of each object. * Removes legacy `ApplyToAllCacheEntries` and changes `ApplyToAllEntries` signature for Deleter->CacheItemHelper change. ## typed_cache.h Adds various "typed" interfaces to the Cache as internal APIs, so that most uses of Cache can use simple type safe code without casting and without explicit deleters, etc. Almost all of the non-test, non-glue code uses of Cache have been migrated. (Follow-up work: CompressedSecondaryCache deserves deeper attention to migrate.) This change expands RocksDB's internal usage of metaprogramming and SFINAE (https://en.cppreference.com/w/cpp/language/sfinae). The existing usages of Cache are divided up at a high level into these new interfaces. See updated existing uses of Cache for examples of how these are used. * PlaceholderCacheInterface - Used for making cache reservations, with entries that have a charge but no value. * BasicTypedCacheInterface<TValue> - Used for primary cache storage of objects of type TValue, which can be cleaned up with std::default_delete<TValue>. The role is provided by TValue::kCacheEntryRole or given in an optional template parameter. * FullTypedCacheInterface<TValue, TCreateContext> - Used for secondary cache compatible storage of objects of type TValue. In addition to BasicTypedCacheInterface constraints, we require TValue::ContentSlice() to return persistable data. This simplifies usage for the normal case of simple secondary cache compatibility (can give you a Slice to the data already in memory). In addition to TCreateContext performing the role of Cache::CreateContext, it is also expected to provide a factory function for creating TValue. * For each of these, there's a "Shared" version (e.g. FullTypedSharedCacheInterface) that holds a shared_ptr to the Cache, rather than assuming external ownership by holding only a raw `Cache*`. These interfaces introduce specific handle types for each interface instantiation, so that it's easy to see what kind of object is controlled by a handle. (Ultimately, this might not be worth the extra complexity, but it seems OK so far.) Note: I attempted to make the cache 'charge' automatically inferred from the cache object type, such as by expecting an ApproximateMemoryUsage() function, but this is not so clean because there are cases where we need to compute the charge ahead of time and don't want to re-compute it. ## block_cache.h This header is essentially the replacement for the old block_like_traits.h. It includes various things to support block cache access with typed_cache.h for block-based table. ## block_based_table_reader.cc Before this change, accessing the block cache here was an awkward mix of static polymorphism (template TBlocklike) and switch-case on a dynamic BlockType value. This change mostly unifies on static polymorphism, relying on minor hacks in block_cache.h to distinguish variants of Block. We still check BlockType in some places (especially for stats, which could be improved in follow-up work) but at least the BlockType is a static constant from the template parameter. (No more awkward partial redundancy between static and dynamic info.) This likely contributes to the overall performance improvement, but hasn't been tested in isolation. The other key source of simplification here is a more unified system of creating block cache objects: for directly populating from primary cache and for promotion from secondary cache. Both use BlockCreateContext, for context and for factory functions. ## block_based_table_builder.cc, cache_dump_load_impl.cc Before this change, warming caches was super ugly code. Both of these source files had switch statements to basically transition from the dynamic BlockType world to the static TBlocklike world. None of that mess is needed anymore as there's a new, untyped WarmInCache function that handles all the details just as promotion from SecondaryCache would. (Fixes `TODO akanksha: Dedup below code` in block_based_table_builder.cc.) ## Everything else Mostly just updating Cache users to use new typed APIs when reasonably possible, or changed Cache APIs when not. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10975 Test Plan: tests updated Performance test setup similar to https://github.com/facebook/rocksdb/issues/10626 (by cache size, LRUCache when not "hyper" for HyperClockCache): 34MB 1thread base.hyper -> kops/s: 0.745 io_bytes/op: 2.52504e+06 miss_ratio: 0.140906 max_rss_mb: 76.4844 34MB 1thread new.hyper -> kops/s: 0.751 io_bytes/op: 2.5123e+06 miss_ratio: 0.140161 max_rss_mb: 79.3594 34MB 1thread base -> kops/s: 0.254 io_bytes/op: 1.36073e+07 miss_ratio: 0.918818 max_rss_mb: 45.9297 34MB 1thread new -> kops/s: 0.252 io_bytes/op: 1.36157e+07 miss_ratio: 0.918999 max_rss_mb: 44.1523 34MB 32thread base.hyper -> kops/s: 7.272 io_bytes/op: 2.88323e+06 miss_ratio: 0.162532 max_rss_mb: 516.602 34MB 32thread new.hyper -> kops/s: 7.214 io_bytes/op: 2.99046e+06 miss_ratio: 0.168818 max_rss_mb: 518.293 34MB 32thread base -> kops/s: 3.528 io_bytes/op: 1.35722e+07 miss_ratio: 0.914691 max_rss_mb: 264.926 34MB 32thread new -> kops/s: 3.604 io_bytes/op: 1.35744e+07 miss_ratio: 0.915054 max_rss_mb: 264.488 233MB 1thread base.hyper -> kops/s: 53.909 io_bytes/op: 2552.35 miss_ratio: 0.0440566 max_rss_mb: 241.984 233MB 1thread new.hyper -> kops/s: 62.792 io_bytes/op: 2549.79 miss_ratio: 0.044043 max_rss_mb: 241.922 233MB 1thread base -> kops/s: 1.197 io_bytes/op: 2.75173e+06 miss_ratio: 0.103093 max_rss_mb: 241.559 233MB 1thread new -> kops/s: 1.199 io_bytes/op: 2.73723e+06 miss_ratio: 0.10305 max_rss_mb: 240.93 233MB 32thread base.hyper -> kops/s: 1298.69 io_bytes/op: 2539.12 miss_ratio: 0.0440307 max_rss_mb: 371.418 233MB 32thread new.hyper -> kops/s: 1421.35 io_bytes/op: 2538.75 miss_ratio: 0.0440307 max_rss_mb: 347.273 233MB 32thread base -> kops/s: 9.693 io_bytes/op: 2.77304e+06 miss_ratio: 0.103745 max_rss_mb: 569.691 233MB 32thread new -> kops/s: 9.75 io_bytes/op: 2.77559e+06 miss_ratio: 0.103798 max_rss_mb: 552.82 1597MB 1thread base.hyper -> kops/s: 58.607 io_bytes/op: 1449.14 miss_ratio: 0.0249324 max_rss_mb: 1583.55 1597MB 1thread new.hyper -> kops/s: 69.6 io_bytes/op: 1434.89 miss_ratio: 0.0247167 max_rss_mb: 1584.02 1597MB 1thread base -> kops/s: 60.478 io_bytes/op: 1421.28 miss_ratio: 0.024452 max_rss_mb: 1589.45 1597MB 1thread new -> kops/s: 63.973 io_bytes/op: 1416.07 miss_ratio: 0.0243766 max_rss_mb: 1589.24 1597MB 32thread base.hyper -> kops/s: 1436.2 io_bytes/op: 1357.93 miss_ratio: 0.0235353 max_rss_mb: 1692.92 1597MB 32thread new.hyper -> kops/s: 1605.03 io_bytes/op: 1358.04 miss_ratio: 0.023538 max_rss_mb: 1702.78 1597MB 32thread base -> kops/s: 280.059 io_bytes/op: 1350.34 miss_ratio: 0.023289 max_rss_mb: 1675.36 1597MB 32thread new -> kops/s: 283.125 io_bytes/op: 1351.05 miss_ratio: 0.0232797 max_rss_mb: 1703.83 Almost uniformly improving over base revision, especially for hot paths with HyperClockCache, up to 12% higher throughput seen (1597MB, 32thread, hyper). The improvement for that is likely coming from much simplified code for providing context for secondary cache promotion (CreateCallback/CreateContext), and possibly from less branching in block_based_table_reader. And likely a small improvement from not reconstituting key for DeleterFn. Reviewed By: anand1976 Differential Revision: D42417818 Pulled By: pdillinger fbshipit-source-id: f86bfdd584dce27c028b151ba56818ad14f7a432
2023-01-11 22:20:40 +00:00
cache_.Release(handle);
Multi file concurrency in MultiGet using coroutines and async IO (#9968) Summary: This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code. A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest. TODO: 1. Figure out how to build it in CircleCI (requires some dependencies to be installed) 2. Do some stress testing with coroutines enabled No regression in synchronous MultiGet between this branch and main - ``` ./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics ``` Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)``` Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)``` More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file. 1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) - No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)``` Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)``` 2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file - No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)``` Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)``` 3. Single thread CPU bound workload with ~2 key overlap/file - No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)``` Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)``` 4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file - No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ``` Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968 Reviewed By: akankshamahajan15 Differential Revision: D36348563 Pulled By: anand1976 fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
2022-05-19 22:36:27 +00:00
}
CO_RETURN s;
}
} // namespace ROCKSDB_NAMESPACE
#endif