rocksdb/table/unique_id.cc
Peter Dillinger 86a1e3e0e7 Derive cache keys from SST unique IDs (#10394)
Summary:
... so that cache keys can be derived from DB manifest data
before reading the file from storage--so that every part of the file
can potentially go in a persistent cache.

See updated comments in cache_key.cc for technical details. Importantly,
the new cache key encoding uses some fancy but efficient math to pack
data into the cache key without depending on the sizes of the various
pieces. This simplifies some existing code creating cache keys, like
cache warming before the file size is known.

This should provide us an essentially permanent mapping between SST
unique IDs and base cache keys, with the ability to "upgrade" SST
unique IDs (and thus cache keys) with new SST format_versions.

These cache keys are of similar, perhaps indistinguishable quality to
the previous generation. Before this change (see "corrected" days
between collision):

```
./cache_bench -stress_cache_key -sck_keep_bits=43
18 collisions after 2 x 90 days, est 10 days between (1.15292e+19 corrected)
```

After this change (keep 43 bits, up through 50, to validate "trajectory"
is ok on "corrected" days between collision):
```
19 collisions after 3 x 90 days, est 14.2105 days between (1.63836e+19 corrected)
16 collisions after 5 x 90 days, est 28.125 days between (1.6213e+19 corrected)
15 collisions after 7 x 90 days, est 42 days between (1.21057e+19 corrected)
15 collisions after 17 x 90 days, est 102 days between (1.46997e+19 corrected)
15 collisions after 49 x 90 days, est 294 days between (2.11849e+19 corrected)
15 collisions after 62 x 90 days, est 372 days between (1.34027e+19 corrected)
15 collisions after 53 x 90 days, est 318 days between (5.72858e+18 corrected)
15 collisions after 309 x 90 days, est 1854 days between (1.66994e+19 corrected)
```

However, the change does modify (probably weaken) the "guaranteed unique" promise from this

> SST files generated in a single process are guaranteed to have unique cache keys, unless/until number session ids * max file number = 2**86

to this (see https://github.com/facebook/rocksdb/issues/10388)

> With the DB id limitation, we only have nice guaranteed unique cache keys for files generated in a single process until biggest session_id_counter and offset_in_file reach combined 64 bits

I don't think this is a practical concern, though.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/10394

Test Plan: unit tests updated, see simulation results above

Reviewed By: jay-zhuang

Differential Revision: D38667529

Pulled By: pdillinger

fbshipit-source-id: 49af3fe7f47e5b61162809a78b76c769fd519fba
2022-08-12 13:49:49 -07:00

224 lines
7.3 KiB
C++

// Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
// This source code is licensed under both the GPLv2 (found in the
// COPYING file in the root directory) and Apache 2.0 License
// (found in the LICENSE.Apache file in the root directory).
#include <cstdint>
#include "table/unique_id_impl.h"
#include "util/coding_lean.h"
#include "util/hash.h"
#include "util/string_util.h"
namespace ROCKSDB_NAMESPACE {
std::string EncodeSessionId(uint64_t upper, uint64_t lower) {
std::string db_session_id(20U, '\0');
char *buf = &db_session_id[0];
// Preserving `lower` is slightly tricky. 36^12 is slightly more than
// 62 bits, so we use 12 chars plus the bottom two bits of one more.
// (A tiny fraction of 20 digit strings go unused.)
uint64_t a = (upper << 2) | (lower >> 62);
uint64_t b = lower & (UINT64_MAX >> 2);
PutBaseChars<36>(&buf, 8, a, /*uppercase*/ true);
PutBaseChars<36>(&buf, 12, b, /*uppercase*/ true);
assert(buf == &db_session_id.back() + 1);
return db_session_id;
}
Status DecodeSessionId(const std::string &db_session_id, uint64_t *upper,
uint64_t *lower) {
const size_t len = db_session_id.size();
if (len == 0) {
return Status::NotSupported("Missing db_session_id");
}
// Anything from 13 to 24 chars is reasonable. We don't have to limit to
// exactly 20.
if (len < 13) {
return Status::NotSupported("Too short db_session_id");
}
if (len > 24) {
return Status::NotSupported("Too long db_session_id");
}
uint64_t a = 0, b = 0;
const char *buf = &db_session_id.front();
bool success = ParseBaseChars<36>(&buf, len - 12U, &a);
if (!success) {
return Status::NotSupported("Bad digit in db_session_id");
}
success = ParseBaseChars<36>(&buf, 12U, &b);
if (!success) {
return Status::NotSupported("Bad digit in db_session_id");
}
assert(buf == &db_session_id.back() + 1);
*upper = a >> 2;
*lower = (b & (UINT64_MAX >> 2)) | (a << 62);
return Status::OK();
}
Status GetSstInternalUniqueId(const std::string &db_id,
const std::string &db_session_id,
uint64_t file_number, UniqueIdPtr out,
bool force) {
if (!force) {
if (db_id.empty()) {
return Status::NotSupported("Missing db_id");
}
if (file_number == 0) {
return Status::NotSupported("Missing or bad file number");
}
if (db_session_id.empty()) {
return Status::NotSupported("Missing db_session_id");
}
}
uint64_t session_upper = 0; // Assignment to appease clang-analyze
uint64_t session_lower = 0; // Assignment to appease clang-analyze
{
Status s = DecodeSessionId(db_session_id, &session_upper, &session_lower);
if (!s.ok()) {
if (!force) {
return s;
} else {
// A reasonable fallback in case malformed
Hash2x64(db_session_id.data(), db_session_id.size(), &session_upper,
&session_lower);
if (session_lower == 0) {
session_lower = session_upper | 1;
}
}
}
}
// Exactly preserve session lower to ensure that session ids generated
// during the same process lifetime are guaranteed unique.
// DBImpl also guarantees (in recent versions) that this is not zero,
// so that we can guarantee unique ID is never all zeros. (Can't assert
// that here because of testing and old versions.)
// We put this first in anticipation of matching a small-ish set of cache
// key prefixes to cover entries relevant to any DB.
out.ptr[0] = session_lower;
// Hash the session upper (~39 bits entropy) and DB id (120+ bits entropy)
// for very high global uniqueness entropy.
// (It is possible that many DBs descended from one common DB id are copied
// around and proliferate, in which case session id is critical, but it is
// more common for different DBs to have different DB ids.)
uint64_t db_a, db_b;
Hash2x64(db_id.data(), db_id.size(), session_upper, &db_a, &db_b);
// Xor in file number for guaranteed uniqueness by file number for a given
// session and DB id. (Xor slightly better than + here. See
// https://github.com/pdillinger/unique_id )
out.ptr[1] = db_a ^ file_number;
// Extra (optional) global uniqueness
if (out.extended) {
out.ptr[2] = db_b;
}
return Status::OK();
}
namespace {
// For InternalUniqueIdToExternal / ExternalUniqueIdToInternal we want all
// zeros in first 128 bits to map to itself, so that excluding zero in
// internal IDs (session_lower != 0 above) does the same for external IDs.
// These values are meaningless except for making that work.
constexpr uint64_t kHiOffsetForZero = 17391078804906429400U;
constexpr uint64_t kLoOffsetForZero = 6417269962128484497U;
} // namespace
void InternalUniqueIdToExternal(UniqueIdPtr in_out) {
uint64_t hi, lo;
BijectiveHash2x64(in_out.ptr[1] + kHiOffsetForZero,
in_out.ptr[0] + kLoOffsetForZero, &hi, &lo);
in_out.ptr[0] = lo;
in_out.ptr[1] = hi;
if (in_out.extended) {
in_out.ptr[2] += lo + hi;
}
}
void ExternalUniqueIdToInternal(UniqueIdPtr in_out) {
uint64_t lo = in_out.ptr[0];
uint64_t hi = in_out.ptr[1];
if (in_out.extended) {
in_out.ptr[2] -= lo + hi;
}
BijectiveUnhash2x64(hi, lo, &hi, &lo);
in_out.ptr[0] = lo - kLoOffsetForZero;
in_out.ptr[1] = hi - kHiOffsetForZero;
}
std::string EncodeUniqueIdBytes(UniqueIdPtr in) {
std::string ret(in.extended ? 24U : 16U, '\0');
EncodeFixed64(&ret[0], in.ptr[0]);
EncodeFixed64(&ret[8], in.ptr[1]);
if (in.extended) {
EncodeFixed64(&ret[16], in.ptr[2]);
}
return ret;
}
Status DecodeUniqueIdBytes(const std::string &unique_id, UniqueIdPtr out) {
if (unique_id.size() != (out.extended ? 24 : 16)) {
return Status::NotSupported("Not a valid unique_id");
}
const char *buf = &unique_id.front();
out.ptr[0] = DecodeFixed64(&buf[0]);
out.ptr[1] = DecodeFixed64(&buf[8]);
if (out.extended) {
out.ptr[2] = DecodeFixed64(&buf[16]);
}
return Status::OK();
}
template <typename ID>
Status GetUniqueIdFromTablePropertiesHelper(const TableProperties &props,
std::string *out_id) {
ID tmp{};
Status s = GetSstInternalUniqueId(props.db_id, props.db_session_id,
props.orig_file_number, &tmp);
if (s.ok()) {
InternalUniqueIdToExternal(&tmp);
*out_id = EncodeUniqueIdBytes(&tmp);
} else {
out_id->clear();
}
return s;
}
Status GetExtendedUniqueIdFromTableProperties(const TableProperties &props,
std::string *out_id) {
return GetUniqueIdFromTablePropertiesHelper<UniqueId64x3>(props, out_id);
}
Status GetUniqueIdFromTableProperties(const TableProperties &props,
std::string *out_id) {
return GetUniqueIdFromTablePropertiesHelper<UniqueId64x2>(props, out_id);
}
std::string UniqueIdToHumanString(const std::string &id) {
// Not so efficient, but that's OK
std::string str = Slice(id).ToString(/*hex*/ true);
for (size_t i = 16; i < str.size(); i += 17) {
str.insert(i, "-");
}
return str;
}
std::string InternalUniqueIdToHumanString(UniqueIdPtr in) {
std::string str = "{";
str += std::to_string(in.ptr[0]);
str += ",";
str += std::to_string(in.ptr[1]);
if (in.extended) {
str += ",";
str += std::to_string(in.ptr[2]);
}
str += "}";
return str;
}
} // namespace ROCKSDB_NAMESPACE