|
|
|
|
// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
|
|
|
|
|
// This source code is licensed under both the GPLv2 (found in the
|
|
|
|
|
// COPYING file in the root directory) and Apache 2.0 License
|
|
|
|
|
// (found in the LICENSE.Apache file in the root directory).
|
|
|
|
|
//
|
|
|
|
|
#include "rocksdb/statistics.h"
|
|
|
|
|
|
|
|
|
|
#include <algorithm>
|
|
|
|
|
#include <cinttypes>
|
|
|
|
|
#include <cstdio>
|
|
|
|
|
|
|
|
|
|
#include "monitoring/statistics_impl.h"
|
|
|
|
|
#include "rocksdb/convenience.h"
|
|
|
|
|
#include "rocksdb/utilities/customizable_util.h"
|
|
|
|
|
#include "rocksdb/utilities/options_type.h"
|
|
|
|
|
#include "util/string_util.h"
|
|
|
|
|
|
|
|
|
|
namespace ROCKSDB_NAMESPACE {
|
|
|
|
|
|
|
|
|
|
// The order of items listed in Tickers should be the same as
|
|
|
|
|
// the order listed in TickersNameMap
|
|
|
|
|
const std::vector<std::pair<Tickers, std::string>> TickersNameMap = {
|
|
|
|
|
{BLOCK_CACHE_MISS, "rocksdb.block.cache.miss"},
|
|
|
|
|
{BLOCK_CACHE_HIT, "rocksdb.block.cache.hit"},
|
|
|
|
|
{BLOCK_CACHE_ADD, "rocksdb.block.cache.add"},
|
|
|
|
|
{BLOCK_CACHE_ADD_FAILURES, "rocksdb.block.cache.add.failures"},
|
|
|
|
|
{BLOCK_CACHE_INDEX_MISS, "rocksdb.block.cache.index.miss"},
|
|
|
|
|
{BLOCK_CACHE_INDEX_HIT, "rocksdb.block.cache.index.hit"},
|
|
|
|
|
{BLOCK_CACHE_INDEX_ADD, "rocksdb.block.cache.index.add"},
|
|
|
|
|
{BLOCK_CACHE_INDEX_BYTES_INSERT, "rocksdb.block.cache.index.bytes.insert"},
|
|
|
|
|
{BLOCK_CACHE_FILTER_MISS, "rocksdb.block.cache.filter.miss"},
|
|
|
|
|
{BLOCK_CACHE_FILTER_HIT, "rocksdb.block.cache.filter.hit"},
|
|
|
|
|
{BLOCK_CACHE_FILTER_ADD, "rocksdb.block.cache.filter.add"},
|
|
|
|
|
{BLOCK_CACHE_FILTER_BYTES_INSERT,
|
|
|
|
|
"rocksdb.block.cache.filter.bytes.insert"},
|
|
|
|
|
{BLOCK_CACHE_DATA_MISS, "rocksdb.block.cache.data.miss"},
|
|
|
|
|
{BLOCK_CACHE_DATA_HIT, "rocksdb.block.cache.data.hit"},
|
|
|
|
|
{BLOCK_CACHE_DATA_ADD, "rocksdb.block.cache.data.add"},
|
|
|
|
|
{BLOCK_CACHE_DATA_BYTES_INSERT, "rocksdb.block.cache.data.bytes.insert"},
|
|
|
|
|
{BLOCK_CACHE_BYTES_READ, "rocksdb.block.cache.bytes.read"},
|
|
|
|
|
{BLOCK_CACHE_BYTES_WRITE, "rocksdb.block.cache.bytes.write"},
|
|
|
|
|
{BLOOM_FILTER_USEFUL, "rocksdb.bloom.filter.useful"},
|
|
|
|
|
{BLOOM_FILTER_FULL_POSITIVE, "rocksdb.bloom.filter.full.positive"},
|
|
|
|
|
{BLOOM_FILTER_FULL_TRUE_POSITIVE,
|
|
|
|
|
"rocksdb.bloom.filter.full.true.positive"},
|
|
|
|
|
{PERSISTENT_CACHE_HIT, "rocksdb.persistent.cache.hit"},
|
|
|
|
|
{PERSISTENT_CACHE_MISS, "rocksdb.persistent.cache.miss"},
|
|
|
|
|
{SIM_BLOCK_CACHE_HIT, "rocksdb.sim.block.cache.hit"},
|
|
|
|
|
{SIM_BLOCK_CACHE_MISS, "rocksdb.sim.block.cache.miss"},
|
|
|
|
|
{MEMTABLE_HIT, "rocksdb.memtable.hit"},
|
|
|
|
|
{MEMTABLE_MISS, "rocksdb.memtable.miss"},
|
|
|
|
|
{GET_HIT_L0, "rocksdb.l0.hit"},
|
|
|
|
|
{GET_HIT_L1, "rocksdb.l1.hit"},
|
|
|
|
|
{GET_HIT_L2_AND_UP, "rocksdb.l2andup.hit"},
|
|
|
|
|
{COMPACTION_KEY_DROP_NEWER_ENTRY, "rocksdb.compaction.key.drop.new"},
|
|
|
|
|
{COMPACTION_KEY_DROP_OBSOLETE, "rocksdb.compaction.key.drop.obsolete"},
|
|
|
|
|
{COMPACTION_KEY_DROP_RANGE_DEL, "rocksdb.compaction.key.drop.range_del"},
|
|
|
|
|
{COMPACTION_KEY_DROP_USER, "rocksdb.compaction.key.drop.user"},
|
|
|
|
|
{COMPACTION_RANGE_DEL_DROP_OBSOLETE,
|
|
|
|
|
"rocksdb.compaction.range_del.drop.obsolete"},
|
|
|
|
|
{COMPACTION_OPTIMIZED_DEL_DROP_OBSOLETE,
|
|
|
|
|
"rocksdb.compaction.optimized.del.drop.obsolete"},
|
|
|
|
|
{COMPACTION_CANCELLED, "rocksdb.compaction.cancelled"},
|
|
|
|
|
{NUMBER_KEYS_WRITTEN, "rocksdb.number.keys.written"},
|
|
|
|
|
{NUMBER_KEYS_READ, "rocksdb.number.keys.read"},
|
|
|
|
|
{NUMBER_KEYS_UPDATED, "rocksdb.number.keys.updated"},
|
|
|
|
|
{BYTES_WRITTEN, "rocksdb.bytes.written"},
|
|
|
|
|
{BYTES_READ, "rocksdb.bytes.read"},
|
|
|
|
|
{NUMBER_DB_SEEK, "rocksdb.number.db.seek"},
|
|
|
|
|
{NUMBER_DB_NEXT, "rocksdb.number.db.next"},
|
|
|
|
|
{NUMBER_DB_PREV, "rocksdb.number.db.prev"},
|
|
|
|
|
{NUMBER_DB_SEEK_FOUND, "rocksdb.number.db.seek.found"},
|
|
|
|
|
{NUMBER_DB_NEXT_FOUND, "rocksdb.number.db.next.found"},
|
|
|
|
|
{NUMBER_DB_PREV_FOUND, "rocksdb.number.db.prev.found"},
|
|
|
|
|
{ITER_BYTES_READ, "rocksdb.db.iter.bytes.read"},
|
|
|
|
|
{NO_FILE_OPENS, "rocksdb.no.file.opens"},
|
|
|
|
|
{NO_FILE_ERRORS, "rocksdb.no.file.errors"},
|
|
|
|
|
{STALL_MICROS, "rocksdb.stall.micros"},
|
|
|
|
|
{DB_MUTEX_WAIT_MICROS, "rocksdb.db.mutex.wait.micros"},
|
|
|
|
|
{NUMBER_MULTIGET_CALLS, "rocksdb.number.multiget.get"},
|
|
|
|
|
{NUMBER_MULTIGET_KEYS_READ, "rocksdb.number.multiget.keys.read"},
|
|
|
|
|
{NUMBER_MULTIGET_BYTES_READ, "rocksdb.number.multiget.bytes.read"},
|
|
|
|
|
{NUMBER_MERGE_FAILURES, "rocksdb.number.merge.failures"},
|
|
|
|
|
{BLOOM_FILTER_PREFIX_CHECKED, "rocksdb.bloom.filter.prefix.checked"},
|
|
|
|
|
{BLOOM_FILTER_PREFIX_USEFUL, "rocksdb.bloom.filter.prefix.useful"},
|
|
|
|
|
{NUMBER_OF_RESEEKS_IN_ITERATION, "rocksdb.number.reseeks.iteration"},
|
|
|
|
|
{GET_UPDATES_SINCE_CALLS, "rocksdb.getupdatessince.calls"},
|
|
|
|
|
{WAL_FILE_SYNCED, "rocksdb.wal.synced"},
|
|
|
|
|
{WAL_FILE_BYTES, "rocksdb.wal.bytes"},
|
|
|
|
|
{WRITE_DONE_BY_SELF, "rocksdb.write.self"},
|
|
|
|
|
{WRITE_DONE_BY_OTHER, "rocksdb.write.other"},
|
|
|
|
|
{WRITE_WITH_WAL, "rocksdb.write.wal"},
|
|
|
|
|
{COMPACT_READ_BYTES, "rocksdb.compact.read.bytes"},
|
|
|
|
|
{COMPACT_WRITE_BYTES, "rocksdb.compact.write.bytes"},
|
|
|
|
|
{FLUSH_WRITE_BYTES, "rocksdb.flush.write.bytes"},
|
|
|
|
|
{COMPACT_READ_BYTES_MARKED, "rocksdb.compact.read.marked.bytes"},
|
|
|
|
|
{COMPACT_READ_BYTES_PERIODIC, "rocksdb.compact.read.periodic.bytes"},
|
|
|
|
|
{COMPACT_READ_BYTES_TTL, "rocksdb.compact.read.ttl.bytes"},
|
|
|
|
|
{COMPACT_WRITE_BYTES_MARKED, "rocksdb.compact.write.marked.bytes"},
|
|
|
|
|
{COMPACT_WRITE_BYTES_PERIODIC, "rocksdb.compact.write.periodic.bytes"},
|
|
|
|
|
{COMPACT_WRITE_BYTES_TTL, "rocksdb.compact.write.ttl.bytes"},
|
|
|
|
|
{NUMBER_DIRECT_LOAD_TABLE_PROPERTIES,
|
|
|
|
|
"rocksdb.number.direct.load.table.properties"},
|
|
|
|
|
{NUMBER_SUPERVERSION_ACQUIRES, "rocksdb.number.superversion_acquires"},
|
|
|
|
|
{NUMBER_SUPERVERSION_RELEASES, "rocksdb.number.superversion_releases"},
|
|
|
|
|
{NUMBER_SUPERVERSION_CLEANUPS, "rocksdb.number.superversion_cleanups"},
|
|
|
|
|
{NUMBER_BLOCK_COMPRESSED, "rocksdb.number.block.compressed"},
|
|
|
|
|
{NUMBER_BLOCK_DECOMPRESSED, "rocksdb.number.block.decompressed"},
|
|
|
|
|
{NUMBER_BLOCK_NOT_COMPRESSED, "rocksdb.number.block.not_compressed"},
|
|
|
|
|
{MERGE_OPERATION_TOTAL_TIME, "rocksdb.merge.operation.time.nanos"},
|
|
|
|
|
{FILTER_OPERATION_TOTAL_TIME, "rocksdb.filter.operation.time.nanos"},
|
|
|
|
|
{ROW_CACHE_HIT, "rocksdb.row.cache.hit"},
|
|
|
|
|
{ROW_CACHE_MISS, "rocksdb.row.cache.miss"},
|
|
|
|
|
{READ_AMP_ESTIMATE_USEFUL_BYTES, "rocksdb.read.amp.estimate.useful.bytes"},
|
|
|
|
|
{READ_AMP_TOTAL_READ_BYTES, "rocksdb.read.amp.total.read.bytes"},
|
|
|
|
|
{NUMBER_RATE_LIMITER_DRAINS, "rocksdb.number.rate_limiter.drains"},
|
|
|
|
|
{NUMBER_ITER_SKIP, "rocksdb.number.iter.skip"},
|
|
|
|
|
{BLOB_DB_NUM_PUT, "rocksdb.blobdb.num.put"},
|
|
|
|
|
{BLOB_DB_NUM_WRITE, "rocksdb.blobdb.num.write"},
|
|
|
|
|
{BLOB_DB_NUM_GET, "rocksdb.blobdb.num.get"},
|
|
|
|
|
{BLOB_DB_NUM_MULTIGET, "rocksdb.blobdb.num.multiget"},
|
|
|
|
|
{BLOB_DB_NUM_SEEK, "rocksdb.blobdb.num.seek"},
|
|
|
|
|
{BLOB_DB_NUM_NEXT, "rocksdb.blobdb.num.next"},
|
|
|
|
|
{BLOB_DB_NUM_PREV, "rocksdb.blobdb.num.prev"},
|
|
|
|
|
{BLOB_DB_NUM_KEYS_WRITTEN, "rocksdb.blobdb.num.keys.written"},
|
|
|
|
|
{BLOB_DB_NUM_KEYS_READ, "rocksdb.blobdb.num.keys.read"},
|
|
|
|
|
{BLOB_DB_BYTES_WRITTEN, "rocksdb.blobdb.bytes.written"},
|
|
|
|
|
{BLOB_DB_BYTES_READ, "rocksdb.blobdb.bytes.read"},
|
|
|
|
|
{BLOB_DB_WRITE_INLINED, "rocksdb.blobdb.write.inlined"},
|
|
|
|
|
{BLOB_DB_WRITE_INLINED_TTL, "rocksdb.blobdb.write.inlined.ttl"},
|
|
|
|
|
{BLOB_DB_WRITE_BLOB, "rocksdb.blobdb.write.blob"},
|
|
|
|
|
{BLOB_DB_WRITE_BLOB_TTL, "rocksdb.blobdb.write.blob.ttl"},
|
|
|
|
|
{BLOB_DB_BLOB_FILE_BYTES_WRITTEN, "rocksdb.blobdb.blob.file.bytes.written"},
|
|
|
|
|
{BLOB_DB_BLOB_FILE_BYTES_READ, "rocksdb.blobdb.blob.file.bytes.read"},
|
|
|
|
|
{BLOB_DB_BLOB_FILE_SYNCED, "rocksdb.blobdb.blob.file.synced"},
|
|
|
|
|
{BLOB_DB_BLOB_INDEX_EXPIRED_COUNT,
|
|
|
|
|
"rocksdb.blobdb.blob.index.expired.count"},
|
|
|
|
|
{BLOB_DB_BLOB_INDEX_EXPIRED_SIZE, "rocksdb.blobdb.blob.index.expired.size"},
|
|
|
|
|
{BLOB_DB_BLOB_INDEX_EVICTED_COUNT,
|
|
|
|
|
"rocksdb.blobdb.blob.index.evicted.count"},
|
|
|
|
|
{BLOB_DB_BLOB_INDEX_EVICTED_SIZE, "rocksdb.blobdb.blob.index.evicted.size"},
|
|
|
|
|
{BLOB_DB_GC_NUM_FILES, "rocksdb.blobdb.gc.num.files"},
|
|
|
|
|
{BLOB_DB_GC_NUM_NEW_FILES, "rocksdb.blobdb.gc.num.new.files"},
|
|
|
|
|
{BLOB_DB_GC_FAILURES, "rocksdb.blobdb.gc.failures"},
|
|
|
|
|
{BLOB_DB_GC_NUM_KEYS_RELOCATED, "rocksdb.blobdb.gc.num.keys.relocated"},
|
|
|
|
|
{BLOB_DB_GC_BYTES_RELOCATED, "rocksdb.blobdb.gc.bytes.relocated"},
|
|
|
|
|
{BLOB_DB_FIFO_NUM_FILES_EVICTED, "rocksdb.blobdb.fifo.num.files.evicted"},
|
|
|
|
|
{BLOB_DB_FIFO_NUM_KEYS_EVICTED, "rocksdb.blobdb.fifo.num.keys.evicted"},
|
|
|
|
|
{BLOB_DB_FIFO_BYTES_EVICTED, "rocksdb.blobdb.fifo.bytes.evicted"},
|
|
|
|
|
{TXN_PREPARE_MUTEX_OVERHEAD, "rocksdb.txn.overhead.mutex.prepare"},
|
|
|
|
|
{TXN_OLD_COMMIT_MAP_MUTEX_OVERHEAD,
|
|
|
|
|
"rocksdb.txn.overhead.mutex.old.commit.map"},
|
|
|
|
|
{TXN_DUPLICATE_KEY_OVERHEAD, "rocksdb.txn.overhead.duplicate.key"},
|
|
|
|
|
{TXN_SNAPSHOT_MUTEX_OVERHEAD, "rocksdb.txn.overhead.mutex.snapshot"},
|
|
|
|
|
{TXN_GET_TRY_AGAIN, "rocksdb.txn.get.tryagain"},
|
|
|
|
|
{NUMBER_MULTIGET_KEYS_FOUND, "rocksdb.number.multiget.keys.found"},
|
|
|
|
|
{NO_ITERATOR_CREATED, "rocksdb.num.iterator.created"},
|
|
|
|
|
{NO_ITERATOR_DELETED, "rocksdb.num.iterator.deleted"},
|
|
|
|
|
{BLOCK_CACHE_COMPRESSION_DICT_MISS,
|
|
|
|
|
"rocksdb.block.cache.compression.dict.miss"},
|
|
|
|
|
{BLOCK_CACHE_COMPRESSION_DICT_HIT,
|
|
|
|
|
"rocksdb.block.cache.compression.dict.hit"},
|
|
|
|
|
{BLOCK_CACHE_COMPRESSION_DICT_ADD,
|
|
|
|
|
"rocksdb.block.cache.compression.dict.add"},
|
|
|
|
|
{BLOCK_CACHE_COMPRESSION_DICT_BYTES_INSERT,
|
|
|
|
|
"rocksdb.block.cache.compression.dict.bytes.insert"},
|
|
|
|
|
{BLOCK_CACHE_ADD_REDUNDANT, "rocksdb.block.cache.add.redundant"},
|
|
|
|
|
{BLOCK_CACHE_INDEX_ADD_REDUNDANT,
|
|
|
|
|
"rocksdb.block.cache.index.add.redundant"},
|
|
|
|
|
{BLOCK_CACHE_FILTER_ADD_REDUNDANT,
|
|
|
|
|
"rocksdb.block.cache.filter.add.redundant"},
|
|
|
|
|
{BLOCK_CACHE_DATA_ADD_REDUNDANT, "rocksdb.block.cache.data.add.redundant"},
|
|
|
|
|
{BLOCK_CACHE_COMPRESSION_DICT_ADD_REDUNDANT,
|
|
|
|
|
"rocksdb.block.cache.compression.dict.add.redundant"},
|
|
|
|
|
{FILES_MARKED_TRASH, "rocksdb.files.marked.trash"},
|
|
|
|
|
{FILES_DELETED_IMMEDIATELY, "rocksdb.files.deleted.immediately"},
|
|
|
|
|
{ERROR_HANDLER_BG_ERROR_COUNT, "rocksdb.error.handler.bg.errro.count"},
|
|
|
|
|
{ERROR_HANDLER_BG_IO_ERROR_COUNT,
|
|
|
|
|
"rocksdb.error.handler.bg.io.errro.count"},
|
|
|
|
|
{ERROR_HANDLER_BG_RETRYABLE_IO_ERROR_COUNT,
|
|
|
|
|
"rocksdb.error.handler.bg.retryable.io.errro.count"},
|
|
|
|
|
{ERROR_HANDLER_AUTORESUME_COUNT, "rocksdb.error.handler.autoresume.count"},
|
|
|
|
|
{ERROR_HANDLER_AUTORESUME_RETRY_TOTAL_COUNT,
|
|
|
|
|
"rocksdb.error.handler.autoresume.retry.total.count"},
|
|
|
|
|
{ERROR_HANDLER_AUTORESUME_SUCCESS_COUNT,
|
|
|
|
|
"rocksdb.error.handler.autoresume.success.count"},
|
|
|
|
|
{MEMTABLE_PAYLOAD_BYTES_AT_FLUSH,
|
|
|
|
|
"rocksdb.memtable.payload.bytes.at.flush"},
|
|
|
|
|
{MEMTABLE_GARBAGE_BYTES_AT_FLUSH,
|
|
|
|
|
"rocksdb.memtable.garbage.bytes.at.flush"},
|
|
|
|
|
{SECONDARY_CACHE_HITS, "rocksdb.secondary.cache.hits"},
|
|
|
|
|
{VERIFY_CHECKSUM_READ_BYTES, "rocksdb.verify_checksum.read.bytes"},
|
|
|
|
|
{BACKUP_READ_BYTES, "rocksdb.backup.read.bytes"},
|
|
|
|
|
{BACKUP_WRITE_BYTES, "rocksdb.backup.write.bytes"},
|
|
|
|
|
{REMOTE_COMPACT_READ_BYTES, "rocksdb.remote.compact.read.bytes"},
|
|
|
|
|
{REMOTE_COMPACT_WRITE_BYTES, "rocksdb.remote.compact.write.bytes"},
|
|
|
|
|
{HOT_FILE_READ_BYTES, "rocksdb.hot.file.read.bytes"},
|
|
|
|
|
{WARM_FILE_READ_BYTES, "rocksdb.warm.file.read.bytes"},
|
|
|
|
|
{COLD_FILE_READ_BYTES, "rocksdb.cold.file.read.bytes"},
|
|
|
|
|
{HOT_FILE_READ_COUNT, "rocksdb.hot.file.read.count"},
|
|
|
|
|
{WARM_FILE_READ_COUNT, "rocksdb.warm.file.read.count"},
|
|
|
|
|
{COLD_FILE_READ_COUNT, "rocksdb.cold.file.read.count"},
|
|
|
|
|
{LAST_LEVEL_READ_BYTES, "rocksdb.last.level.read.bytes"},
|
|
|
|
|
{LAST_LEVEL_READ_COUNT, "rocksdb.last.level.read.count"},
|
|
|
|
|
{NON_LAST_LEVEL_READ_BYTES, "rocksdb.non.last.level.read.bytes"},
|
|
|
|
|
{NON_LAST_LEVEL_READ_COUNT, "rocksdb.non.last.level.read.count"},
|
|
|
|
|
{BLOCK_CHECKSUM_COMPUTE_COUNT, "rocksdb.block.checksum.compute.count"},
|
|
|
|
|
{BLOCK_CHECKSUM_MISMATCH_COUNT, "rocksdb.block.checksum.mismatch.count"},
|
|
|
|
|
{MULTIGET_COROUTINE_COUNT, "rocksdb.multiget.coroutine.count"},
|
|
|
|
|
{BLOB_DB_CACHE_MISS, "rocksdb.blobdb.cache.miss"},
|
|
|
|
|
{BLOB_DB_CACHE_HIT, "rocksdb.blobdb.cache.hit"},
|
|
|
|
|
{BLOB_DB_CACHE_ADD, "rocksdb.blobdb.cache.add"},
|
|
|
|
|
{BLOB_DB_CACHE_ADD_FAILURES, "rocksdb.blobdb.cache.add.failures"},
|
|
|
|
|
{BLOB_DB_CACHE_BYTES_READ, "rocksdb.blobdb.cache.bytes.read"},
|
|
|
|
|
{BLOB_DB_CACHE_BYTES_WRITE, "rocksdb.blobdb.cache.bytes.write"},
|
|
|
|
|
{READ_ASYNC_MICROS, "rocksdb.read.async.micros"},
|
|
|
|
|
{ASYNC_READ_ERROR_COUNT, "rocksdb.async.read.error.count"},
|
|
|
|
|
{SECONDARY_CACHE_FILTER_HITS, "rocksdb.secondary.cache.filter.hits"},
|
|
|
|
|
{SECONDARY_CACHE_INDEX_HITS, "rocksdb.secondary.cache.index.hits"},
|
Add new stat rocksdb.table.open.prefetch.tail.read.bytes, rocksdb.table.open.prefetch.tail.{miss|hit} (#11265)
Summary:
**Context/Summary:**
We are adding new stats to measure behavior of prefetched tail size and look up into this buffer
The stat collection is done in FilePrefetchBuffer but only for prefetched tail buffer during table open for now using FilePrefetchBuffer enum. It's cleaner than the alternative of implementing in upper-level call places of FilePrefetchBuffer for table open. It also has the benefit of extensible to other types of FilePrefetchBuffer if needed. See db bench for perf regression concern.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11265
Test Plan:
**- Piggyback on existing test**
**- rocksdb.table.open.prefetch.tail.miss is harder to UT so I manually set prefetch tail read bytes to be small and run db bench.**
```
./db_bench -db=/tmp/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=false -compression_type=none -bloom_bits=3 -use_direct_reads=true
```
```
rocksdb.table.open.prefetch.tail.read.bytes P50 : 4096.000000 P95 : 4096.000000 P99 : 4096.000000 P100 : 4096.000000 COUNT : 225 SUM : 921600
rocksdb.table.open.prefetch.tail.miss COUNT : 91
rocksdb.table.open.prefetch.tail.hit COUNT : 1034
```
**- No perf regression observed in db_bench**
SETUP command: create same db with ~900 files for pre-change/post-change.
```
./db_bench -db=/tmp/testdb -benchmarks="fillseq" -key_size=32 -value_size=512 -num=500000 -write_buffer_size=655360 -disable_auto_compactions=true -target_file_size_base=16777216 -compression_type=none
```
TEST command 60 runs or til convergence: as suggested by anand1976 and akankshamahajan15, vary `seek_nexts` and `async_io` in testing.
```
./db_bench -use_existing_db=true -db=/tmp/testdb -statistics=false -cache_size=0 -cache_index_and_filter_blocks=false -benchmarks=seekrandom[-X60] -num=50000 -seek_nexts={10, 500, 1000} -async_io={0|1} -use_direct_reads=true
```
async io = 0, direct io read = true
| seek_nexts = 10, 30 runs | seek_nexts = 500, 12 runs | seek_nexts = 1000, 6 runs
-- | -- | -- | --
pre-post change | 4776 (± 28) ops/sec; 24.8 (± 0.1) MB/sec | 288 (± 1) ops/sec; 74.8 (± 0.4) MB/sec | 145 (± 4) ops/sec; 75.6 (± 2.2) MB/sec
post-change | 4790 (± 32) ops/sec; 24.9 (± 0.2) MB/sec | 288 (± 3) ops/sec; 74.7 (± 0.8) MB/sec | 143 (± 3) ops/sec; 74.5 (± 1.6) MB/sec
async io = 1, direct io read = true
| seek_nexts = 10, 54 runs | seek_nexts = 500, 6 runs | seek_nexts = 1000, 4 runs
-- | -- | -- | --
pre-post change | 3350 (± 36) ops/sec; 17.4 (± 0.2) MB/sec | 264 (± 0) ops/sec; 68.7 (± 0.2) MB/sec | 138 (± 1) ops/sec; 71.8 (± 1.0) MB/sec
post-change | 3358 (± 27) ops/sec; 17.4 (± 0.1) MB/sec | 263 (± 2) ops/sec; 68.3 (± 0.8) MB/sec | 139 (± 1) ops/sec; 72.6 (± 0.6) MB/sec
Reviewed By: ajkr
Differential Revision: D43781467
Pulled By: hx235
fbshipit-source-id: a706a18472a8edb2b952bac3af40eec803537f2a
2 years ago
|
|
|
{SECONDARY_CACHE_DATA_HITS, "rocksdb.secondary.cache.data.hits"},
|
|
|
|
|
{TABLE_OPEN_PREFETCH_TAIL_MISS, "rocksdb.table.open.prefetch.tail.miss"},
|
|
|
|
|
{TABLE_OPEN_PREFETCH_TAIL_HIT, "rocksdb.table.open.prefetch.tail.hit"},
|
|
|
|
|
{TIMESTAMP_FILTER_TABLE_CHECKED, "rocksdb.timestamp.filter.table.checked"},
|
|
|
|
|
{TIMESTAMP_FILTER_TABLE_FILTERED,
|
|
|
|
|
"rocksdb.timestamp.filter.table.filtered"},
|
|
|
|
|
{BYTES_COMPRESSED_FROM, "rocksdb.bytes.compressed.from"},
|
|
|
|
|
{BYTES_COMPRESSED_TO, "rocksdb.bytes.compressed.to"},
|
|
|
|
|
{BYTES_COMPRESSION_BYPASSED, "rocksdb.bytes.compression_bypassed"},
|
|
|
|
|
{BYTES_COMPRESSION_REJECTED, "rocksdb.bytes.compression.rejected"},
|
|
|
|
|
{NUMBER_BLOCK_COMPRESSION_BYPASSED,
|
|
|
|
|
"rocksdb.number.block_compression_bypassed"},
|
|
|
|
|
{NUMBER_BLOCK_COMPRESSION_REJECTED,
|
|
|
|
|
"rocksdb.number.block_compression_rejected"},
|
|
|
|
|
{BYTES_DECOMPRESSED_FROM, "rocksdb.bytes.decompressed.from"},
|
|
|
|
|
{BYTES_DECOMPRESSED_TO, "rocksdb.bytes.decompressed.to"},
|
Add new stat rocksdb.table.open.prefetch.tail.read.bytes, rocksdb.table.open.prefetch.tail.{miss|hit} (#11265)
Summary:
**Context/Summary:**
We are adding new stats to measure behavior of prefetched tail size and look up into this buffer
The stat collection is done in FilePrefetchBuffer but only for prefetched tail buffer during table open for now using FilePrefetchBuffer enum. It's cleaner than the alternative of implementing in upper-level call places of FilePrefetchBuffer for table open. It also has the benefit of extensible to other types of FilePrefetchBuffer if needed. See db bench for perf regression concern.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11265
Test Plan:
**- Piggyback on existing test**
**- rocksdb.table.open.prefetch.tail.miss is harder to UT so I manually set prefetch tail read bytes to be small and run db bench.**
```
./db_bench -db=/tmp/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=false -compression_type=none -bloom_bits=3 -use_direct_reads=true
```
```
rocksdb.table.open.prefetch.tail.read.bytes P50 : 4096.000000 P95 : 4096.000000 P99 : 4096.000000 P100 : 4096.000000 COUNT : 225 SUM : 921600
rocksdb.table.open.prefetch.tail.miss COUNT : 91
rocksdb.table.open.prefetch.tail.hit COUNT : 1034
```
**- No perf regression observed in db_bench**
SETUP command: create same db with ~900 files for pre-change/post-change.
```
./db_bench -db=/tmp/testdb -benchmarks="fillseq" -key_size=32 -value_size=512 -num=500000 -write_buffer_size=655360 -disable_auto_compactions=true -target_file_size_base=16777216 -compression_type=none
```
TEST command 60 runs or til convergence: as suggested by anand1976 and akankshamahajan15, vary `seek_nexts` and `async_io` in testing.
```
./db_bench -use_existing_db=true -db=/tmp/testdb -statistics=false -cache_size=0 -cache_index_and_filter_blocks=false -benchmarks=seekrandom[-X60] -num=50000 -seek_nexts={10, 500, 1000} -async_io={0|1} -use_direct_reads=true
```
async io = 0, direct io read = true
| seek_nexts = 10, 30 runs | seek_nexts = 500, 12 runs | seek_nexts = 1000, 6 runs
-- | -- | -- | --
pre-post change | 4776 (± 28) ops/sec; 24.8 (± 0.1) MB/sec | 288 (± 1) ops/sec; 74.8 (± 0.4) MB/sec | 145 (± 4) ops/sec; 75.6 (± 2.2) MB/sec
post-change | 4790 (± 32) ops/sec; 24.9 (± 0.2) MB/sec | 288 (± 3) ops/sec; 74.7 (± 0.8) MB/sec | 143 (± 3) ops/sec; 74.5 (± 1.6) MB/sec
async io = 1, direct io read = true
| seek_nexts = 10, 54 runs | seek_nexts = 500, 6 runs | seek_nexts = 1000, 4 runs
-- | -- | -- | --
pre-post change | 3350 (± 36) ops/sec; 17.4 (± 0.2) MB/sec | 264 (± 0) ops/sec; 68.7 (± 0.2) MB/sec | 138 (± 1) ops/sec; 71.8 (± 1.0) MB/sec
post-change | 3358 (± 27) ops/sec; 17.4 (± 0.1) MB/sec | 263 (± 2) ops/sec; 68.3 (± 0.8) MB/sec | 139 (± 1) ops/sec; 72.6 (± 0.6) MB/sec
Reviewed By: ajkr
Differential Revision: D43781467
Pulled By: hx235
fbshipit-source-id: a706a18472a8edb2b952bac3af40eec803537f2a
2 years ago
|
|
|
};
|
|
|
|
|
|
|
|
|
|
const std::vector<std::pair<Histograms, std::string>> HistogramsNameMap = {
|
|
|
|
|
{DB_GET, "rocksdb.db.get.micros"},
|
|
|
|
|
{DB_WRITE, "rocksdb.db.write.micros"},
|
|
|
|
|
{COMPACTION_TIME, "rocksdb.compaction.times.micros"},
|
|
|
|
|
{COMPACTION_CPU_TIME, "rocksdb.compaction.times.cpu_micros"},
|
|
|
|
|
{SUBCOMPACTION_SETUP_TIME, "rocksdb.subcompaction.setup.times.micros"},
|
|
|
|
|
{TABLE_SYNC_MICROS, "rocksdb.table.sync.micros"},
|
|
|
|
|
{COMPACTION_OUTFILE_SYNC_MICROS, "rocksdb.compaction.outfile.sync.micros"},
|
|
|
|
|
{WAL_FILE_SYNC_MICROS, "rocksdb.wal.file.sync.micros"},
|
|
|
|
|
{MANIFEST_FILE_SYNC_MICROS, "rocksdb.manifest.file.sync.micros"},
|
|
|
|
|
{TABLE_OPEN_IO_MICROS, "rocksdb.table.open.io.micros"},
|
|
|
|
|
{DB_MULTIGET, "rocksdb.db.multiget.micros"},
|
|
|
|
|
{READ_BLOCK_COMPACTION_MICROS, "rocksdb.read.block.compaction.micros"},
|
|
|
|
|
{READ_BLOCK_GET_MICROS, "rocksdb.read.block.get.micros"},
|
|
|
|
|
{WRITE_RAW_BLOCK_MICROS, "rocksdb.write.raw.block.micros"},
|
|
|
|
|
{NUM_FILES_IN_SINGLE_COMPACTION, "rocksdb.numfiles.in.singlecompaction"},
|
|
|
|
|
{DB_SEEK, "rocksdb.db.seek.micros"},
|
|
|
|
|
{WRITE_STALL, "rocksdb.db.write.stall"},
|
|
|
|
|
{SST_READ_MICROS, "rocksdb.sst.read.micros"},
|
|
|
|
|
{FILE_READ_FLUSH_MICROS, "rocksdb.file.read.flush.micros"},
|
|
|
|
|
{FILE_READ_COMPACTION_MICROS, "rocksdb.file.read.compaction.micros"},
|
|
|
|
|
{NUM_SUBCOMPACTIONS_SCHEDULED, "rocksdb.num.subcompactions.scheduled"},
|
|
|
|
|
{BYTES_PER_READ, "rocksdb.bytes.per.read"},
|
|
|
|
|
{BYTES_PER_WRITE, "rocksdb.bytes.per.write"},
|
|
|
|
|
{BYTES_PER_MULTIGET, "rocksdb.bytes.per.multiget"},
|
|
|
|
|
{BYTES_COMPRESSED, "rocksdb.bytes.compressed"},
|
|
|
|
|
{BYTES_DECOMPRESSED, "rocksdb.bytes.decompressed"},
|
|
|
|
|
{COMPRESSION_TIMES_NANOS, "rocksdb.compression.times.nanos"},
|
|
|
|
|
{DECOMPRESSION_TIMES_NANOS, "rocksdb.decompression.times.nanos"},
|
|
|
|
|
{READ_NUM_MERGE_OPERANDS, "rocksdb.read.num.merge_operands"},
|
|
|
|
|
{BLOB_DB_KEY_SIZE, "rocksdb.blobdb.key.size"},
|
|
|
|
|
{BLOB_DB_VALUE_SIZE, "rocksdb.blobdb.value.size"},
|
|
|
|
|
{BLOB_DB_WRITE_MICROS, "rocksdb.blobdb.write.micros"},
|
|
|
|
|
{BLOB_DB_GET_MICROS, "rocksdb.blobdb.get.micros"},
|
|
|
|
|
{BLOB_DB_MULTIGET_MICROS, "rocksdb.blobdb.multiget.micros"},
|
|
|
|
|
{BLOB_DB_SEEK_MICROS, "rocksdb.blobdb.seek.micros"},
|
|
|
|
|
{BLOB_DB_NEXT_MICROS, "rocksdb.blobdb.next.micros"},
|
|
|
|
|
{BLOB_DB_PREV_MICROS, "rocksdb.blobdb.prev.micros"},
|
|
|
|
|
{BLOB_DB_BLOB_FILE_WRITE_MICROS, "rocksdb.blobdb.blob.file.write.micros"},
|
|
|
|
|
{BLOB_DB_BLOB_FILE_READ_MICROS, "rocksdb.blobdb.blob.file.read.micros"},
|
|
|
|
|
{BLOB_DB_BLOB_FILE_SYNC_MICROS, "rocksdb.blobdb.blob.file.sync.micros"},
|
|
|
|
|
{BLOB_DB_COMPRESSION_MICROS, "rocksdb.blobdb.compression.micros"},
|
|
|
|
|
{BLOB_DB_DECOMPRESSION_MICROS, "rocksdb.blobdb.decompression.micros"},
|
|
|
|
|
{FLUSH_TIME, "rocksdb.db.flush.micros"},
|
Introduce a new MultiGet batching implementation (#5011)
Summary:
This PR introduces a new MultiGet() API, with the underlying implementation grouping keys based on SST file and batching lookups in a file. The reason for the new API is twofold - the definition allows callers to allocate storage for status and values on stack instead of std::vector, as well as return values as PinnableSlices in order to avoid copying, and it keeps the original MultiGet() implementation intact while we experiment with batching.
Batching is useful when there is some spatial locality to the keys being queries, as well as larger batch sizes. The main benefits are due to -
1. Fewer function calls, especially to BlockBasedTableReader::MultiGet() and FullFilterBlockReader::KeysMayMatch()
2. Bloom filter cachelines can be prefetched, hiding the cache miss latency
The next step is to optimize the binary searches in the level_storage_info, index blocks and data blocks, since we could reduce the number of key comparisons if the keys are relatively close to each other. The batching optimizations also need to be extended to other formats, such as PlainTable and filter formats. This also needs to be added to db_stress.
Benchmark results from db_bench for various batch size/locality of reference combinations are given below. Locality was simulated by offsetting the keys in a batch by a stride length. Each SST file is about 8.6MB uncompressed and key/value size is 16/100 uncompressed. To focus on the cpu benefit of batching, the runs were single threaded and bound to the same cpu to eliminate interference from other system events. The results show a 10-25% improvement in micros/op from smaller to larger batch sizes (4 - 32).
Batch Sizes
1 | 2 | 4 | 8 | 16 | 32
Random pattern (Stride length 0)
4.158 | 4.109 | 4.026 | 4.05 | 4.1 | 4.074 - Get
4.438 | 4.302 | 4.165 | 4.122 | 4.096 | 4.075 - MultiGet (no batching)
4.461 | 4.256 | 4.277 | 4.11 | 4.182 | 4.14 - MultiGet (w/ batching)
Good locality (Stride length 16)
4.048 | 3.659 | 3.248 | 2.99 | 2.84 | 2.753
4.429 | 3.728 | 3.406 | 3.053 | 2.911 | 2.781
4.452 | 3.45 | 2.833 | 2.451 | 2.233 | 2.135
Good locality (Stride length 256)
4.066 | 3.786 | 3.581 | 3.447 | 3.415 | 3.232
4.406 | 4.005 | 3.644 | 3.49 | 3.381 | 3.268
4.393 | 3.649 | 3.186 | 2.882 | 2.676 | 2.62
Medium locality (Stride length 4096)
4.012 | 3.922 | 3.768 | 3.61 | 3.582 | 3.555
4.364 | 4.057 | 3.791 | 3.65 | 3.57 | 3.465
4.479 | 3.758 | 3.316 | 3.077 | 2.959 | 2.891
dbbench command used (on a DB with 4 levels, 12 million keys)-
TEST_TMPDIR=/dev/shm numactl -C 10 ./db_bench.tmp -use_existing_db=true -benchmarks="readseq,multireadrandom" -write_buffer_size=4194304 -target_file_size_base=4194304 -max_bytes_for_level_base=16777216 -num=12000000 -reads=12000000 -duration=90 -threads=1 -compression_type=none -cache_size=4194304000 -batch_size=32 -disable_auto_compactions=true -bloom_bits=10 -cache_index_and_filter_blocks=true -pin_l0_filter_and_index_blocks_in_cache=true -multiread_batched=true -multiread_stride=4
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5011
Differential Revision: D14348703
Pulled By: anand1976
fbshipit-source-id: 774406dab3776d979c809522a67bedac6c17f84b
6 years ago
|
|
|
{SST_BATCH_SIZE, "rocksdb.sst.batch.size"},
|
|
|
|
|
{NUM_INDEX_AND_FILTER_BLOCKS_READ_PER_LEVEL,
|
|
|
|
|
"rocksdb.num.index.and.filter.blocks.read.per.level"},
|
|
|
|
|
{NUM_SST_READ_PER_LEVEL, "rocksdb.num.sst.read.per.level"},
|
|
|
|
|
{ERROR_HANDLER_AUTORESUME_RETRY_COUNT,
|
|
|
|
|
"rocksdb.error.handler.autoresume.retry.count"},
|
|
|
|
|
{ASYNC_READ_BYTES, "rocksdb.async.read.bytes"},
|
|
|
|
|
{POLL_WAIT_MICROS, "rocksdb.poll.wait.micros"},
|
|
|
|
|
{PREFETCHED_BYTES_DISCARDED, "rocksdb.prefetched.bytes.discarded"},
|
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
3 years ago
|
|
|
{MULTIGET_IO_BATCH_SIZE, "rocksdb.multiget.io.batch.size"},
|
|
|
|
|
{NUM_LEVEL_READ_PER_MULTIGET, "rocksdb.num.level.read.per.multiget"},
|
|
|
|
|
{ASYNC_PREFETCH_ABORT_MICROS, "rocksdb.async.prefetch.abort.micros"},
|
Add new stat rocksdb.table.open.prefetch.tail.read.bytes, rocksdb.table.open.prefetch.tail.{miss|hit} (#11265)
Summary:
**Context/Summary:**
We are adding new stats to measure behavior of prefetched tail size and look up into this buffer
The stat collection is done in FilePrefetchBuffer but only for prefetched tail buffer during table open for now using FilePrefetchBuffer enum. It's cleaner than the alternative of implementing in upper-level call places of FilePrefetchBuffer for table open. It also has the benefit of extensible to other types of FilePrefetchBuffer if needed. See db bench for perf regression concern.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11265
Test Plan:
**- Piggyback on existing test**
**- rocksdb.table.open.prefetch.tail.miss is harder to UT so I manually set prefetch tail read bytes to be small and run db bench.**
```
./db_bench -db=/tmp/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=false -compression_type=none -bloom_bits=3 -use_direct_reads=true
```
```
rocksdb.table.open.prefetch.tail.read.bytes P50 : 4096.000000 P95 : 4096.000000 P99 : 4096.000000 P100 : 4096.000000 COUNT : 225 SUM : 921600
rocksdb.table.open.prefetch.tail.miss COUNT : 91
rocksdb.table.open.prefetch.tail.hit COUNT : 1034
```
**- No perf regression observed in db_bench**
SETUP command: create same db with ~900 files for pre-change/post-change.
```
./db_bench -db=/tmp/testdb -benchmarks="fillseq" -key_size=32 -value_size=512 -num=500000 -write_buffer_size=655360 -disable_auto_compactions=true -target_file_size_base=16777216 -compression_type=none
```
TEST command 60 runs or til convergence: as suggested by anand1976 and akankshamahajan15, vary `seek_nexts` and `async_io` in testing.
```
./db_bench -use_existing_db=true -db=/tmp/testdb -statistics=false -cache_size=0 -cache_index_and_filter_blocks=false -benchmarks=seekrandom[-X60] -num=50000 -seek_nexts={10, 500, 1000} -async_io={0|1} -use_direct_reads=true
```
async io = 0, direct io read = true
| seek_nexts = 10, 30 runs | seek_nexts = 500, 12 runs | seek_nexts = 1000, 6 runs
-- | -- | -- | --
pre-post change | 4776 (± 28) ops/sec; 24.8 (± 0.1) MB/sec | 288 (± 1) ops/sec; 74.8 (± 0.4) MB/sec | 145 (± 4) ops/sec; 75.6 (± 2.2) MB/sec
post-change | 4790 (± 32) ops/sec; 24.9 (± 0.2) MB/sec | 288 (± 3) ops/sec; 74.7 (± 0.8) MB/sec | 143 (± 3) ops/sec; 74.5 (± 1.6) MB/sec
async io = 1, direct io read = true
| seek_nexts = 10, 54 runs | seek_nexts = 500, 6 runs | seek_nexts = 1000, 4 runs
-- | -- | -- | --
pre-post change | 3350 (± 36) ops/sec; 17.4 (± 0.2) MB/sec | 264 (± 0) ops/sec; 68.7 (± 0.2) MB/sec | 138 (± 1) ops/sec; 71.8 (± 1.0) MB/sec
post-change | 3358 (± 27) ops/sec; 17.4 (± 0.1) MB/sec | 263 (± 2) ops/sec; 68.3 (± 0.8) MB/sec | 139 (± 1) ops/sec; 72.6 (± 0.6) MB/sec
Reviewed By: ajkr
Differential Revision: D43781467
Pulled By: hx235
fbshipit-source-id: a706a18472a8edb2b952bac3af40eec803537f2a
2 years ago
|
|
|
{TABLE_OPEN_PREFETCH_TAIL_READ_BYTES,
|
|
|
|
|
"rocksdb.table.open.prefetch.tail.read.bytes"},
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
std::shared_ptr<Statistics> CreateDBStatistics() {
|
|
|
|
|
return std::make_shared<StatisticsImpl>(nullptr);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int RegisterBuiltinStatistics(ObjectLibrary& library,
|
|
|
|
|
const std::string& /*arg*/) {
|
|
|
|
|
library.AddFactory<Statistics>(
|
|
|
|
|
StatisticsImpl::kClassName(),
|
|
|
|
|
[](const std::string& /*uri*/, std::unique_ptr<Statistics>* guard,
|
|
|
|
|
std::string* /* errmsg */) {
|
|
|
|
|
guard->reset(new StatisticsImpl(nullptr));
|
|
|
|
|
return guard->get();
|
|
|
|
|
});
|
|
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
Status Statistics::CreateFromString(const ConfigOptions& config_options,
|
|
|
|
|
const std::string& id,
|
|
|
|
|
std::shared_ptr<Statistics>* result) {
|
|
|
|
|
static std::once_flag once;
|
|
|
|
|
std::call_once(once, [&]() {
|
|
|
|
|
RegisterBuiltinStatistics(*(ObjectLibrary::Default().get()), "");
|
|
|
|
|
});
|
|
|
|
|
Status s;
|
|
|
|
|
if (id == "" || id == StatisticsImpl::kClassName()) {
|
|
|
|
|
result->reset(new StatisticsImpl(nullptr));
|
|
|
|
|
} else if (id == kNullptrString) {
|
|
|
|
|
result->reset();
|
|
|
|
|
} else {
|
|
|
|
|
s = LoadSharedObject<Statistics>(config_options, id, result);
|
|
|
|
|
}
|
|
|
|
|
return s;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static std::unordered_map<std::string, OptionTypeInfo> stats_type_info = {
|
|
|
|
|
{"inner", OptionTypeInfo::AsCustomSharedPtr<Statistics>(
|
|
|
|
|
0, OptionVerificationType::kByNameAllowFromNull,
|
|
|
|
|
OptionTypeFlags::kCompareNever)},
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
StatisticsImpl::StatisticsImpl(std::shared_ptr<Statistics> stats)
|
|
|
|
|
: stats_(std::move(stats)) {
|
|
|
|
|
RegisterOptions("StatisticsOptions", &stats_, &stats_type_info);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
StatisticsImpl::~StatisticsImpl() {}
|
|
|
|
|
|
|
|
|
|
uint64_t StatisticsImpl::getTickerCount(uint32_t tickerType) const {
|
|
|
|
|
MutexLock lock(&aggregate_lock_);
|
|
|
|
|
return getTickerCountLocked(tickerType);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
uint64_t StatisticsImpl::getTickerCountLocked(uint32_t tickerType) const {
|
|
|
|
|
assert(tickerType < TICKER_ENUM_MAX);
|
|
|
|
|
uint64_t res = 0;
|
|
|
|
|
for (size_t core_idx = 0; core_idx < per_core_stats_.Size(); ++core_idx) {
|
|
|
|
|
res += per_core_stats_.AccessAtCore(core_idx)->tickers_[tickerType];
|
|
|
|
|
}
|
|
|
|
|
return res;
|
Thread-specific histogram statistics
Summary:
To reduce contention for atomics when HistogramStats are shared across
threads, this diff makes them thread-specific so updates are faster. This comes
at the expense of slower reads (much less frequent), which now require merging
all histograms. In this diff,
- Thread-specific HistogramImpl is created upon the thread's first measureTime()
- Thread-specific HistogramImpl are merged and deleted upon thread termination or ThreadLocalPtr destruction, whichever comes first
- getHistogramString() and histogramData() merge all histograms, both thread-specific and previously merged ones
Test Plan:
unit tests, ran db_bench and verified histograms look similar
before:
$ TEST_TMPDIR=/dev/shm/ perf record -g ./db_bench --benchmarks=readwhilewriting --statistics --num=1000000 --use_existing_db --threads=64 --cache_size=250000000 --compression_type=lz4
...
+ 7.63% db_bench db_bench [.] rocksdb::HistogramStat::Add
after:
$ TEST_TMPDIR=/dev/shm/ perf record -g ./db_bench --benchmarks=readwhilewriting --statistics --num=1000000 --use_existing_db --threads=64 --cache_size=250000000 --compression_type=lz4
...
+ 0.98% db_bench db_bench [.] rocksdb::HistogramStat::Add
Reviewers: sdong, MarkCallaghan, kradhakrishnan, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D62649
8 years ago
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void StatisticsImpl::histogramData(uint32_t histogramType,
|
|
|
|
|
HistogramData* const data) const {
|
|
|
|
|
MutexLock lock(&aggregate_lock_);
|
|
|
|
|
getHistogramImplLocked(histogramType)->Data(data);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
std::unique_ptr<HistogramImpl> StatisticsImpl::getHistogramImplLocked(
|
|
|
|
|
uint32_t histogramType) const {
|
|
|
|
|
assert(histogramType < HISTOGRAM_ENUM_MAX);
|
|
|
|
|
std::unique_ptr<HistogramImpl> res_hist(new HistogramImpl());
|
|
|
|
|
for (size_t core_idx = 0; core_idx < per_core_stats_.Size(); ++core_idx) {
|
|
|
|
|
res_hist->Merge(
|
|
|
|
|
per_core_stats_.AccessAtCore(core_idx)->histograms_[histogramType]);
|
|
|
|
|
}
|
|
|
|
|
return res_hist;
|
|
|
|
|
}
|
|
|
|
|
|
Add Statistics.getHistogramString() to print more detailed outputs of a histogram
Summary:
Provide a way for users to know more detailed ditribution of a histogram metrics. Example outputs:
Manually add statement
fprintf(stdout, "%s\n", dbstats->getHistogramString(SST_READ_MICROS).c_str());
Will print out something like:
Count: 989151 Average: 1.7659 StdDev: 1.52
Min: 0.0000 Median: 1.2071 Max: 860.0000
Percentiles: P50: 1.21 P75: 1.70 P99: 5.12 P99.9: 13.67 P99.99: 21.70
------------------------------------------------------
[ 0, 1 ) 390839 39.513% 39.513% ########
[ 1, 2 ) 500918 50.641% 90.154% ##########
[ 2, 3 ) 79358 8.023% 98.177% ##
[ 3, 4 ) 6297 0.637% 98.813%
[ 4, 5 ) 1712 0.173% 98.986%
[ 5, 6 ) 1134 0.115% 99.101%
[ 6, 7 ) 1222 0.124% 99.224%
[ 7, 8 ) 1529 0.155% 99.379%
[ 8, 9 ) 1264 0.128% 99.507%
[ 9, 10 ) 988 0.100% 99.607%
[ 10, 12 ) 1378 0.139% 99.746%
[ 12, 14 ) 1828 0.185% 99.931%
[ 14, 16 ) 410 0.041% 99.972%
[ 16, 18 ) 72 0.007% 99.980%
[ 18, 20 ) 67 0.007% 99.986%
[ 20, 25 ) 106 0.011% 99.997%
[ 25, 30 ) 24 0.002% 99.999%
[ 30, 35 ) 1 0.000% 100.000%
[ 250, 300 ) 2 0.000% 100.000%
[ 300, 350 ) 1 0.000% 100.000%
[ 800, 900 ) 1 0.000% 100.000%
Test Plan: Manually add a print in db_bench and make sure it prints out as expected. Will add some codes to cover the function
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D43611
9 years ago
|
|
|
std::string StatisticsImpl::getHistogramString(uint32_t histogramType) const {
|
|
|
|
|
MutexLock lock(&aggregate_lock_);
|
|
|
|
|
return getHistogramImplLocked(histogramType)->ToString();
|
Thread-specific histogram statistics
Summary:
To reduce contention for atomics when HistogramStats are shared across
threads, this diff makes them thread-specific so updates are faster. This comes
at the expense of slower reads (much less frequent), which now require merging
all histograms. In this diff,
- Thread-specific HistogramImpl is created upon the thread's first measureTime()
- Thread-specific HistogramImpl are merged and deleted upon thread termination or ThreadLocalPtr destruction, whichever comes first
- getHistogramString() and histogramData() merge all histograms, both thread-specific and previously merged ones
Test Plan:
unit tests, ran db_bench and verified histograms look similar
before:
$ TEST_TMPDIR=/dev/shm/ perf record -g ./db_bench --benchmarks=readwhilewriting --statistics --num=1000000 --use_existing_db --threads=64 --cache_size=250000000 --compression_type=lz4
...
+ 7.63% db_bench db_bench [.] rocksdb::HistogramStat::Add
after:
$ TEST_TMPDIR=/dev/shm/ perf record -g ./db_bench --benchmarks=readwhilewriting --statistics --num=1000000 --use_existing_db --threads=64 --cache_size=250000000 --compression_type=lz4
...
+ 0.98% db_bench db_bench [.] rocksdb::HistogramStat::Add
Reviewers: sdong, MarkCallaghan, kradhakrishnan, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D62649
8 years ago
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void StatisticsImpl::setTickerCount(uint32_t tickerType, uint64_t count) {
|
|
|
|
|
{
|
|
|
|
|
MutexLock lock(&aggregate_lock_);
|
|
|
|
|
setTickerCountLocked(tickerType, count);
|
|
|
|
|
}
|
|
|
|
|
if (stats_ && tickerType < TICKER_ENUM_MAX) {
|
|
|
|
|
stats_->setTickerCount(tickerType, count);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void StatisticsImpl::setTickerCountLocked(uint32_t tickerType, uint64_t count) {
|
|
|
|
|
assert(tickerType < TICKER_ENUM_MAX);
|
|
|
|
|
for (size_t core_idx = 0; core_idx < per_core_stats_.Size(); ++core_idx) {
|
|
|
|
|
if (core_idx == 0) {
|
|
|
|
|
per_core_stats_.AccessAtCore(core_idx)->tickers_[tickerType] = count;
|
|
|
|
|
} else {
|
|
|
|
|
per_core_stats_.AccessAtCore(core_idx)->tickers_[tickerType] = 0;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
uint64_t StatisticsImpl::getAndResetTickerCount(uint32_t tickerType) {
|
|
|
|
|
uint64_t sum = 0;
|
|
|
|
|
{
|
|
|
|
|
MutexLock lock(&aggregate_lock_);
|
|
|
|
|
assert(tickerType < TICKER_ENUM_MAX);
|
|
|
|
|
for (size_t core_idx = 0; core_idx < per_core_stats_.Size(); ++core_idx) {
|
|
|
|
|
sum +=
|
|
|
|
|
per_core_stats_.AccessAtCore(core_idx)->tickers_[tickerType].exchange(
|
|
|
|
|
0, std::memory_order_relaxed);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
if (stats_ && tickerType < TICKER_ENUM_MAX) {
|
|
|
|
|
stats_->setTickerCount(tickerType, 0);
|
|
|
|
|
}
|
|
|
|
|
return sum;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void StatisticsImpl::recordTick(uint32_t tickerType, uint64_t count) {
|
|
|
|
|
if (get_stats_level() <= StatsLevel::kExceptTickers) {
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
if (tickerType < TICKER_ENUM_MAX) {
|
|
|
|
|
per_core_stats_.Access()->tickers_[tickerType].fetch_add(
|
|
|
|
|
count, std::memory_order_relaxed);
|
|
|
|
|
if (stats_) {
|
|
|
|
|
stats_->recordTick(tickerType, count);
|
|
|
|
|
}
|
|
|
|
|
} else {
|
|
|
|
|
assert(false);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void StatisticsImpl::recordInHistogram(uint32_t histogramType, uint64_t value) {
|
|
|
|
|
assert(histogramType < HISTOGRAM_ENUM_MAX);
|
|
|
|
|
if (get_stats_level() <= StatsLevel::kExceptHistogramOrTimers) {
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
per_core_stats_.Access()->histograms_[histogramType].Add(value);
|
|
|
|
|
if (stats_ && histogramType < HISTOGRAM_ENUM_MAX) {
|
|
|
|
|
stats_->recordInHistogram(histogramType, value);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
Status StatisticsImpl::Reset() {
|
|
|
|
|
MutexLock lock(&aggregate_lock_);
|
|
|
|
|
for (uint32_t i = 0; i < TICKER_ENUM_MAX; ++i) {
|
|
|
|
|
setTickerCountLocked(i, 0);
|
|
|
|
|
}
|
|
|
|
|
for (uint32_t i = 0; i < HISTOGRAM_ENUM_MAX; ++i) {
|
|
|
|
|
for (size_t core_idx = 0; core_idx < per_core_stats_.Size(); ++core_idx) {
|
|
|
|
|
per_core_stats_.AccessAtCore(core_idx)->histograms_[i].Clear();
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
return Status::OK();
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
namespace {
|
|
|
|
|
|
|
|
|
|
// a buffer size used for temp string buffers
|
|
|
|
|
const int kTmpStrBufferSize = 200;
|
|
|
|
|
|
|
|
|
|
} // namespace
|
|
|
|
|
|
|
|
|
|
std::string StatisticsImpl::ToString() const {
|
|
|
|
|
MutexLock lock(&aggregate_lock_);
|
|
|
|
|
std::string res;
|
|
|
|
|
res.reserve(20000);
|
|
|
|
|
for (const auto& t : TickersNameMap) {
|
|
|
|
|
assert(t.first < TICKER_ENUM_MAX);
|
|
|
|
|
char buffer[kTmpStrBufferSize];
|
|
|
|
|
snprintf(buffer, kTmpStrBufferSize, "%s COUNT : %" PRIu64 "\n",
|
|
|
|
|
t.second.c_str(), getTickerCountLocked(t.first));
|
|
|
|
|
res.append(buffer);
|
|
|
|
|
}
|
|
|
|
|
for (const auto& h : HistogramsNameMap) {
|
|
|
|
|
assert(h.first < HISTOGRAM_ENUM_MAX);
|
|
|
|
|
char buffer[kTmpStrBufferSize];
|
|
|
|
|
HistogramData hData;
|
|
|
|
|
getHistogramImplLocked(h.first)->Data(&hData);
|
|
|
|
|
// don't handle failures - buffer should always be big enough and arguments
|
|
|
|
|
// should be provided correctly
|
|
|
|
|
int ret =
|
|
|
|
|
snprintf(buffer, kTmpStrBufferSize,
|
|
|
|
|
"%s P50 : %f P95 : %f P99 : %f P100 : %f COUNT : %" PRIu64
|
|
|
|
|
" SUM : %" PRIu64 "\n",
|
|
|
|
|
h.second.c_str(), hData.median, hData.percentile95,
|
|
|
|
|
hData.percentile99, hData.max, hData.count, hData.sum);
|
|
|
|
|
if (ret < 0 || ret >= kTmpStrBufferSize) {
|
|
|
|
|
assert(false);
|
|
|
|
|
continue;
|
|
|
|
|
}
|
|
|
|
|
res.append(buffer);
|
|
|
|
|
}
|
|
|
|
|
res.shrink_to_fit();
|
|
|
|
|
return res;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
bool StatisticsImpl::getTickerMap(
|
|
|
|
|
std::map<std::string, uint64_t>* stats_map) const {
|
|
|
|
|
assert(stats_map);
|
|
|
|
|
if (!stats_map) return false;
|
|
|
|
|
stats_map->clear();
|
|
|
|
|
MutexLock lock(&aggregate_lock_);
|
|
|
|
|
for (const auto& t : TickersNameMap) {
|
|
|
|
|
assert(t.first < TICKER_ENUM_MAX);
|
|
|
|
|
(*stats_map)[t.second.c_str()] = getTickerCountLocked(t.first);
|
|
|
|
|
}
|
|
|
|
|
return true;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
bool StatisticsImpl::HistEnabledForType(uint32_t type) const {
|
|
|
|
|
return type < HISTOGRAM_ENUM_MAX;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
} // namespace ROCKSDB_NAMESPACE
|
