mirror of https://github.com/facebook/rocksdb.git
414 Commits
| Author | SHA1 | Message | Date |
|---|---|---|---|
Peter Dillinger
|
d780e7a561 |
Remove `bottommost_temperature` (#12389)
Summary: deprecated option already replaced by `last_level_temperature`. (Keeping recognition of the option in old options files.) Pull Request resolved: https://github.com/facebook/rocksdb/pull/12389 Test Plan: tests updated Reviewed By: jowlyzhang, cbi42 Differential Revision: D54267946 Pulled By: pdillinger fbshipit-source-id: 65c49b15e7394829c1f3b44edd4179d2daff6017 |
|
|
Andrew Kryczka
|
a43481b3d0 |
Decouple `RateLimiter` burst size and refill period (#12379)
Summary: When the rate limiter does not have any waiting requests, the first request to arrive may consume all of the available bandwidth, despite potentially having lower priority than requests that arrive later in the same refill interval. Then, those higher priority requests must wait for a refill. So even in scenarios in which we have an overall bandwidth surplus, the highest priority requests can be sporadically delayed up to a whole refill period. Alone, this isn't necessarily problematic as the refill period is configurable via `refill_period_us` and can be tuned down as needed until the max sporadic delay is tolerable. However, tuning down `refill_period_us` had a side effect of reducing burst size. Some users require a certain burst size to issue optimal I/O sizes to the underlying storage system. To satisfy those users, this PR decouples the refill period from the burst size. That way, the max sporadic delay can be limited without impacting I/O sizes issued to the underlying storage system. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12379 Test Plan: The goal is to show we can now limit the max sporadic delay without impacting compaction's I/O size. The benchmark runs compaction with a large I/O size, while user reads simultaneously run at a low rate that does not consume all of the available bandwidth. The max sporadic delay is measured using the P100 of rocksdb.file.read.get.micros. I just used strace to verify the compaction reads follow `rate_limiter_single_burst_bytes` Setup: `./db_bench -benchmarks=fillrandom,flush -write_buffer_size=67108864 -disable_auto_compactions=true -value_size=256 -num=1048576` Benchmark: `./db_bench -benchmarks=readrandom -use_existing_db=true -num=1048576 -duration=10 -benchmark_read_rate_limit=4096 -rate_limiter_bytes_per_sec=67108864 -rate_limiter_refill_period_us=$refill_micros -rate_limiter_single_burst_bytes=16777216 -rate_limit_bg_reads=true -rate_limit_user_ops=true -statistics=true -cache_size=0 -stats_level=5 -compaction_readahead_size=16777216 -use_direct_reads=true` Results: refill_micros | rocksdb.file.read.get.micros (P100) -- | -- 10000 | 10802 100000 | 100240 1000000 | 922061 For verifying compaction read sizes: `strace -fye pread64 ./db_bench -benchmarks=compact -use_existing_db=true -rate_limiter_bytes_per_sec=67108864 -rate_limiter_refill_period_us=$refill_micros -rate_limiter_single_burst_bytes=16777216 -rate_limit_bg_reads=true -compaction_readahead_size=16777216 -use_direct_reads=true` Reviewed By: hx235 Differential Revision: D54165675 Pulled By: ajkr fbshipit-source-id: c5968486316cbfb7ff8e5b7d75d3589883dd1105 |
|
|
Andrew Kryczka
|
8e29f243c9 |
No filesystem reads during `Merge()` writes (#12365)
Summary: This occasional filesystem read in the write path has caused user pain. It doesn't seem very useful considering it only limits one component's merge chain length, and only helps merge uncached (i.e., infrequently read) values. This PR proposes allowing `max_successive_merges` to be exceeded when the value cannot be read from in-memory components. I included a rollback flag (`strict_max_successive_merges`) just in case. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12365 Test Plan: "rocksdb.block.cache.data.add" is number of data blocks read from filesystem. Since the benchmark is write-only, compaction is disabled, and flush doesn't read data blocks, any nonzero value means the user write issued the read. ``` $ for s in false true; do echo -n "strict_max_successive_merges=$s: " && ./db_bench -value_size=64 -write_buffer_size=131072 -writes=128 -num=1 -benchmarks=mergerandom,flush,mergerandom -merge_operator=stringappend -disable_auto_compactions=true -compression_type=none -strict_max_successive_merges=$s -max_successive_merges=100 -statistics=true |& grep 'block.cache.data.add COUNT' ; done strict_max_successive_merges=false: rocksdb.block.cache.data.add COUNT : 0 strict_max_successive_merges=true: rocksdb.block.cache.data.add COUNT : 1 ``` Reviewed By: hx235 Differential Revision: D53982520 Pulled By: ajkr fbshipit-source-id: e40f761a60bd601f232417ac0058e4a33ee9c0f4 |
|
|
Peter Dillinger
|
54cb9c77d9 |
Prefer static_cast in place of most reinterpret_cast (#12308)
Summary: The following are risks associated with pointer-to-pointer reinterpret_cast: * Can produce the "wrong result" (crash or memory corruption). IIRC, in theory this can happen for any up-cast or down-cast for a non-standard-layout type, though in practice would only happen for multiple inheritance cases (where the base class pointer might be "inside" the derived object). We don't use multiple inheritance a lot, but we do. * Can mask useful compiler errors upon code change, including converting between unrelated pointer types that you are expecting to be related, and converting between pointer and scalar types unintentionally. I can only think of some obscure cases where static_cast could be troublesome when it compiles as a replacement: * Going through `void*` could plausibly cause unnecessary or broken pointer arithmetic. Suppose we have `struct Derived: public Base1, public Base2`. If we have `Derived*` -> `void*` -> `Base2*` -> `Derived*` through reinterpret casts, this could plausibly work (though technical UB) assuming the `Base2*` is not dereferenced. Changing to static cast could introduce breaking pointer arithmetic. * Unnecessary (but safe) pointer arithmetic could arise in a case like `Derived*` -> `Base2*` -> `Derived*` where before the Base2 pointer might not have been dereferenced. This could potentially affect performance. With some light scripting, I tried replacing pointer-to-pointer reinterpret_casts with static_cast and kept the cases that still compile. Most occurrences of reinterpret_cast have successfully been changed (except for java/ and third-party/). 294 changed, 257 remain. A couple of related interventions included here: * Previously Cache::Handle was not actually derived from in the implementations and just used as a `void*` stand-in with reinterpret_cast. Now there is a relationship to allow static_cast. In theory, this could introduce pointer arithmetic (as described above) but is unlikely without multiple inheritance AND non-empty Cache::Handle. * Remove some unnecessary casts to void* as this is allowed to be implicit (for better or worse). Most of the remaining reinterpret_casts are for converting to/from raw bytes of objects. We could consider better idioms for these patterns in follow-up work. I wish there were a way to implement a template variant of static_cast that would only compile if no pointer arithmetic is generated, but best I can tell, this is not possible. AFAIK the best you could do is a dynamic check that the void* conversion after the static cast is unchanged. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12308 Test Plan: existing tests, CI Reviewed By: ltamasi Differential Revision: D53204947 Pulled By: pdillinger fbshipit-source-id: 9de23e618263b0d5b9820f4e15966876888a16e2 |
|
|
Hui Xiao
|
1a885fe730 |
Remove deprecated Options::access_hint_on_compaction_start (#11654)
Summary:
**Context:**
`Options::access_hint_on_compaction_start ` is marked deprecated and now ready to be removed.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11654
Test Plan:
Multiple db_stress runs with pre-PR and post-PR binary randomly to ensure forward/backward compatibility on options
|
|
|
Andrew Kryczka
|
f9d45358ca |
Removed `check_flush_compaction_key_order` (#12311)
Summary: `check_flush_compaction_key_order` option was introduced for the key order checking online validation. It gave users the ability to disable the validation without downgrade in case the validation caused inefficiencies or false positives. Over time this validation has shown to be cheap and correct, so the option to disable it can now be removed. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12311 Reviewed By: cbi42 Differential Revision: D53233379 Pulled By: ajkr fbshipit-source-id: 1384361104021d6e3e580dce2ec123f9f99ce637 |
|
|
Peter Dillinger
|
76c834e441 |
Remove 'virtual' when implied by 'override' (#12319)
Summary:
... to follow modern C++ style / idioms.
Used this hack:
```
for FILE in `cat my_list_of_files`; do perl -pi -e 'BEGIN{undef $/;} s/ virtual( [^;{]* override)/$1/smg' $FILE; done
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/12319
Test Plan: existing tests, CI
Reviewed By: jaykorean
Differential Revision: D53275303
Pulled By: pdillinger
fbshipit-source-id: bc0881af270aa8ef4d0ae4f44c5a6614b6407377
|
|
|
anand76
|
b49f9cdd3c |
Add CompressionOptions to the compressed secondary cache (#12234)
Summary: Add ```CompressionOptions``` to ```CompressedSecondaryCacheOptions``` to allow users to set options such as compression level. It allows performance to be fine tuned. Tests - Run db_bench and verify compression options in the LOG file Pull Request resolved: https://github.com/facebook/rocksdb/pull/12234 Reviewed By: ajkr Differential Revision: D52758133 Pulled By: anand1976 fbshipit-source-id: af849fbffce6f84704387c195d8edba40d9548f6 |
|
|
Andrew Kryczka
|
06dc32ef25 |
internal_repo_rocksdb (435146444452818992) (#12115)
Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/12115 Reviewed By: jowlyzhang Differential Revision: D51745742 Pulled By: ajkr fbshipit-source-id: 67000d07783b413924798dd9c1751da27e119d53 |
|
|
anand76
|
acc078f878 |
Add tiered cache options to db_bench (#12104)
Summary: Add the option to have a 3-tier block cache (uncompressed RAM, compressed RAM, and local flash) in db_bench, as well as specifying secondary cache admission policy. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12104 Reviewed By: ajkr Differential Revision: D51629092 Pulled By: anand1976 fbshipit-source-id: 6a208f853bc85d3d8b437d91cb1b0142d9a99e53 |
|
|
anand76
|
269478ee46 |
Support compressed and local flash secondary cache stacking (#11812)
Summary: This PR implements support for a three tier cache - primary block cache, compressed secondary cache, and a nvm (local flash) secondary cache. This allows more effective utilization of the nvm cache, and minimizes the number of reads from local flash by caching compressed blocks in the compressed secondary cache. The basic design is as follows - 1. A new secondary cache implementation, ```TieredSecondaryCache```, is introduced. It keeps the compressed and nvm secondary caches and manages the movement of blocks between them and the primary block cache. To setup a three tier cache, we allocate a ```CacheWithSecondaryAdapter```, with a ```TieredSecondaryCache``` instance as the secondary cache. 2. The table reader passes both the uncompressed and compressed block to ```FullTypedCacheInterface::InsertFull```, allowing the block cache to optionally store the compressed block. 3. When there's a miss, the block object is constructed and inserted in the primary cache, and the compressed block is inserted into the nvm cache by calling ```InsertSaved```. This avoids the overhead of recompressing the block, as well as avoiding putting more memory pressure on the compressed secondary cache. 4. When there's a hit in the nvm cache, we attempt to insert the block in the compressed secondary cache and the primary cache, subject to the admission policy of those caches (i.e admit on second access). Blocks/items evicted from any tier are simply discarded. We can easily implement additional admission policies if desired. Todo (In a subsequent PR): 1. Add to db_bench and run benchmarks 2. Add to db_stress Pull Request resolved: https://github.com/facebook/rocksdb/pull/11812 Reviewed By: pdillinger Differential Revision: D49461842 Pulled By: anand1976 fbshipit-source-id: b40ac1330ef7cd8c12efa0a3ca75128e602e3a0b |
|
|
Changyu Bi
|
60de713e15 |
Use uint64_t for `compaction_readahead_size` in stress test (#11849)
Summary:
Internal clang check complains: `tools/db_bench_tool.cc:722:43: error: implicit conversion loses integer precision: 'size_t' (aka 'unsigned long') to 'const gflags::int32' (aka 'const int') [-Werror,-Wshorten-64-to-32]
ROCKSDB_NAMESPACE::Options().compaction_readahead_size,`
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11849
Test Plan: `make -C internal_repo_rocksdb/repo -j64 USE_CLANG=1 J=40 check`, I can only repro when using on-demand devserver.
Reviewed By: hx235
Differential Revision: D49344491
Pulled By: cbi42
fbshipit-source-id: 8c2c0bf2a075c3190b8b91f14f64e26ee252f20f
|
|
|
Hui Xiao
|
b050751f76 |
Use default value instead of hard-coded 0 for compaction_readhead_size in db bench (#11831)
Summary: **Context/Summary:** It allows db bench reflect the default behavior of this option. For example, we recently changed its default value. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11831 Test Plan: No code change Reviewed By: cbi42 Differential Revision: D49253690 Pulled By: hx235 fbshipit-source-id: 445d4e54f62b4b538626e301a3014d2f00849d30 |
|
|
Peter Dillinger
|
fe3405e80f |
Automatic table sizing for HyperClockCache (AutoHCC) (#11738)
Summary: This change add an experimental next-generation HyperClockCache (HCC) with automatic sizing of the underlying hash table. Both the existing version (stable) and the new version (experimental for now) of HCC are available depending on whether an estimated average entry charge is provided in HyperClockCacheOptions. Internally, we call the two implementations AutoHyperClockCache (new) and FixedHyperClockCache (existing). The performance characteristics and much of the underlying logic are similar enough that AutoHCC is likely to make FixedHCC obsolete, and so it's best considered an evolution of the same technology or solution rather than an alternative. More specifically, both implementations share essentially the same logic for managing the state of individual entries in the cache, including metadata for reference counting and counting clocks for eviction. This metadata, which I like to call the "low-level HCC protocol," includes a read-write lock on entries, but relaxed consistency requirements on the cache (e.g. allowing rare duplication) means high-level cache operations never wait for these low-level per-entry locks. FixedHCC is fully wait-free. AutoHCC is different in how entries are indexed into an efficient hash table. AutoHCC is "essentially wait-free" as there is no pattern of typical high-level operations on a large cache that can lead to one thread waiting on another to complete some work, though it can happen in some unusual/unlucky cases, or atypical uses such as erasing specific cache keys. Table growth and entry reclamation is more complex in AutoHCC compared to FixedHCC, so uses some localized locking to manage that. AutoHCC uses linear hashing to grow the table as needed, with low latency and to a precise size. AutoHCC depends on anonymous mmap support from the OS (currently verified working on Linux, MacOS, and Windows) to allow the array underlying a hash table to grow in place without wasting resident memory on space reserved but unused. AutoHCC uses a form of chaining while FixedHCC uses open addressing and double hashing. More specifics: * In developing this PR, a rare availability bug (minor) was noticed in the existing HCC implementation of Release()+erase_if_last_ref, which is now inherited into AutoHCC. Fixing this without a performance regression will not be simple, so is left for follow-up work. * Some existing unit tests required adjustment of operational parameters or conditions to work with the new behaviors of AutoHCC. A number of bugs were found and fixed in the validation process, including getting unit tests in good working order. * Added an option to cache_bench, `-degenerate_hash_bits` for correctness stress testing described below. For this, the tool uses the reverse-engineered hash function for HCC to generate keys in which the specified number of hash bits, in critical positions, have a fixed value. Essentially each degenerate hash bit will half the number of chain heads utilized and double the average chain length. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11738 Test Plan: unit tests updated, and already added to db crash test. Also ## Correctness The code includes generous assertions to check for unexpected states, especially at destruction time, so should be able to detect critical concurrency bugs. Less serious "availability bugs" in which cache data is hidden or cleanly lost are more difficult to detect, but also less scary for data correctness (as long as performance is good and the design is sound). In average operation, the structure is extremely low stress and low contention (see next section) so stressing the corner case logic requires artificially stressing the operating conditions. First, we keep the structure small to increase the number of threads hitting the same chain or entry, and just one cache shard. Second, we artificially degrade the hashing so that chains are much longer than typical, using the new `-degenerate_hash_bits` option to cache_bench. Third, we re-create the structure from scratch frequently in order to exercise the Grow logic repeatedly and to get the benefit of the consistency checks in the structure's destructor in debug builds. For cache_bench this also means disabling the single-threaded "populate cache" step (normally used for steady state performance testing). And of course use many more threads than cores to have many preemptions. An effective test for working out bugs was this (using debug build of course): ``` while ./cache_bench -cache_type=auto_hyper_clock_cache -histograms=0 -cache_size=8000000 -threads=100 -populate_cache=0 -ops_per_thread=10000 -degenerate_hash_bits=6 -num_shard_bits=0; do :; done ``` Or even smaller cases. This setup has around 27 utilized chains, with around 35 entries each, and yield-waits more than 1 million times per second (very high contention; see next section). I have let this run for hours searching for any lingering issues. I've also run cache_bench under ASAN, UBSAN, and TSAN. ## Essentially wait free There is a counter for number of yield() calls when one thread is waiting on another. When we pre-populate the structure in a single thread, ``` ./cache_bench -cache_type=auto_hyper_clock_cache -histograms=0 -populate_cache=1 -ops_per_thread=200000 2>&1 | grep Yield ``` We see something on the order of 1 yield call per second across 16 threads, even when we load the system other other jobs (parallel compilation). With -populate_cache=0, there are more yield opportunities with parallel table growth. On an otherwise unloaded system, we still see very small (single digit) yield counts, with a chance of getting into the thousands, and getting into 10s of thousands per second during table growth phase if the system is loaded with other jobs. However, I am not worried about this if performance is still good (see next section). ## Overall performance Although cache_bench initially suggested performance very close to FixedHCC, there was a very noticeable performance hit under a db_bench setup like used in validating https://github.com/facebook/rocksdb/issues/10626. Much of the difference has been reduced by optimizing Lookup with a "naive" pass that will almost always find entries quickly, and only falling back to the careful Lookup algorithm when not found in the first pass. Setups (chosen to be sensitive to block cache performance), and compiled with USE_CLANG=1 JEMALLOC=1 PORTABLE=0 DEBUG_LEVEL=0: ``` TEST_TMPDIR=/dev/shm base/db_bench -benchmarks=fillrandom -num=30000000 -disable_wal=1 -bloom_bits=16 ``` ### No regression on FixedHCC Running before & after builds at the same time on a 48 core machine. ``` TEST_TMPDIR=/dev/shm /usr/bin/time ./db_bench -benchmarks=readrandom[-X10],block_cache_entry_stats,cache_report_problems -readonly -num=30000000 -bloom_bits=16 -cache_index_and_filter_blocks=1 -cache_size=610000000 -duration 20 -threads=24 -cache_type=fixed_hyper_clock_cache -seed=1234 ``` Before: readrandom [AVG 10 runs] : 847234 (± 8150) ops/sec; 59.2 (± 0.6) MB/sec 703MB max RSS After: readrandom [AVG 10 runs] : 851021 (± 7929) ops/sec; 59.5 (± 0.6) MB/sec 706MB max RSS Probably no material difference. ### Single-threaded performance Using `[-X2]` and `-threads=1` and `-duration=30`, running all three at the same time: lru_cache: 55100 ops/sec, then 55862 ops/sec (627MB max RSS) fixed_hyper_clock_cache: 60496 ops/sec, then 61231 ops/sec (626MB max RSS) auto_hyper_clock_cache: 47560 ops/sec, then 56081 ops/sec (626MB max RSS) So AutoHCC has more ramp-up cost in the first pass as the cache grows to the appropriate size. (In single-threaded operation, the parallelizability and per-op low latency of table growth is overall slower.) However, once up to size, its performance is comparable to LRUCache. FixedHCC's lean operations still win overall when a good estimate is available. If we look at HCC table stats, we can see that this configuration is not favorable to AutoHCC (and I have verified that other memory sizes do not yield substantially different results, until shards are under-sized for the full filters): FixedHCC: Slot occupancy stats: Overall 47% (124991/262144), Min/Max/Window = 28%/64%/500, MaxRun{Pos/Neg} = 17/22 AutoHCC: Slot occupancy stats: Overall 59% (125781/209682), Min/Max/Window = 43%/82%/500, MaxRun{Pos/Neg} = 76/16 Head occupancy stats: Overall 43% (92259/209682), Min/Max/Window = 24%/74%/500, MaxRun{Pos/Neg} = 19/26 Entries at home count: 53350 FixedHCC configuration is relatively good for speed, and not ideal for space utilization. As is typical, AutoHCC has tighter control on metadata usage (209682 x 64 bytes rather than 262144 x 64 bytes), and the higher load factor is slightly worse for speed. LRUCache also has more metadata usage, at 199680 x 96 bytes of tracked metadata (plus roughly another 10% of that untracked in the head pointers), and that metadata is subject to fragmentation. ### Parallel performance, high hit rate Now using `[-X10]` and `-threads=10`, all three at the same time lru_cache: [AVG 10 runs] : 263629 (± 1425) ops/sec; 18.4 (± 0.1) MB/sec 655MB max RSS, 97.1% cache hit rate fixed_hyper_clock_cache: [AVG 10 runs] : 479590 (± 8114) ops/sec; 33.5 (± 0.6) MB/sec 651MB max RSS, 97.1% cache hit rate auto_hyper_clock_cache: [AVG 10 runs] : 418687 (± 5915) ops/sec; 29.3 (± 0.4) MB/sec 657MB max RSS, 97.1% cache hit rate Even with just 10-way parallelism for each cache (though 30+/48 cores busy overall), LRUCache is already showing performance degradation, while AutoHCC is in the neighborhood of FixedHCC. And that brings us to the question of how AutoHCC holds up under extreme parallelism, so now independent runs with `-threads=100` (overloading 48 cores). lru_cache: 438613 ops/sec, 827MB max RSS fixed_hyper_clock_cache: 1651310 ops/sec, 812MB max RSS auto_hyper_clock_cache: 1505875 ops/sec, 821MB max RSS (Yield count: 1089 over 30s) Clearly, AutoHCC holds up extremely well under extreme parallelism, even closing some of the modest performance gap with FixedHCC. ### Parallel performance, low hit rate To get down to roughly 50% cache hit rate, we use `-cache_index_and_filter_blocks=0 -cache_size=1650000000` with `-threads=10`. Here the extra cost of running counting clock eviction, especially on the chains of AutoHCC, are evident, especially with the lower contention of cache_index_and_filter_blocks=0: lru_cache: 725231 ops/sec, 1770MB max RSS, 51.3% hit rate fixed_hyper_clock_cache: 638620 ops/sec, 1765MB max RSS, 50.2% hit rate auto_hyper_clock_cache: 541018 ops/sec, 1777MB max RSS, 50.8% hit rate Reviewed By: jowlyzhang Differential Revision: D48784755 Pulled By: pdillinger fbshipit-source-id: e79813dc087474ac427637dd282a14fa3011a6e4 |
|
|
Akanksha Mahajan
|
6353c6e2fb |
Add new experimental ReadOption auto_readahead_size to db_bench and db_stress (#11729)
Summary: Same as title Pull Request resolved: https://github.com/facebook/rocksdb/pull/11729 Test Plan: make crash_test -j32 Reviewed By: anand1976 Differential Revision: D48534820 Pulled By: akankshamahajan15 fbshipit-source-id: 3a2a28af98dfad164b82ddaaf9fddb94c53a652e |
|
|
Jay Huh
|
4fa2c01719 |
Replace existing waitforcompaction with new WaitForCompact API in db_bench_tool (#11727)
Summary: As the new API to wait for compaction is available (https://github.com/facebook/rocksdb/issues/11436), we can now replace the existing logic of waiting in db_bench_tool with the new API. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11727 Test Plan: ``` ./db_bench --benchmarks="fillrandom,compactall,waitforcompaction,readrandom" ``` **Before change** ``` Set seed to 1692635571470041 because --seed was 0 Initializing RocksDB Options from the specified file Initializing RocksDB Options from command-line flags Integrated BlobDB: blob cache disabled RocksDB: version 8.6.0 Date: Mon Aug 21 09:33:40 2023 CPU: 80 * Intel(R) Xeon(R) Gold 6138 CPU @ 2.00GHz CPUCache: 28160 KB Keys: 16 bytes each (+ 0 bytes user-defined timestamp) Values: 100 bytes each (50 bytes after compression) Entries: 1000000 Prefix: 0 bytes Keys per prefix: 0 RawSize: 110.6 MB (estimated) FileSize: 62.9 MB (estimated) Write rate: 0 bytes/second Read rate: 0 ops/second Compression: Snappy Compression sampling rate: 0 Memtablerep: SkipListFactory Perf Level: 1 WARNING: Optimization is disabled: benchmarks unnecessarily slow WARNING: Assertions are enabled; benchmarks unnecessarily slow ------------------------------------------------ Initializing RocksDB Options from the specified file Initializing RocksDB Options from command-line flags Integrated BlobDB: blob cache disabled DB path: [/tmp/rocksdbtest-226125/dbbench] fillrandom : 51.826 micros/op 19295 ops/sec 51.826 seconds 1000000 operations; 2.1 MB/s waitforcompaction(/tmp/rocksdbtest-226125/dbbench): started waitforcompaction(/tmp/rocksdbtest-226125/dbbench): finished waitforcompaction(/tmp/rocksdbtest-226125/dbbench): started waitforcompaction(/tmp/rocksdbtest-226125/dbbench): finished DB path: [/tmp/rocksdbtest-226125/dbbench] readrandom : 39.042 micros/op 25613 ops/sec 39.042 seconds 1000000 operations; 1.8 MB/s (632886 of 1000000 found) ``` **After change** ``` Set seed to 1692636574431745 because --seed was 0 Initializing RocksDB Options from the specified file Initializing RocksDB Options from command-line flags Integrated BlobDB: blob cache disabled RocksDB: version 8.6.0 Date: Mon Aug 21 09:49:34 2023 CPU: 80 * Intel(R) Xeon(R) Gold 6138 CPU @ 2.00GHz CPUCache: 28160 KB Keys: 16 bytes each (+ 0 bytes user-defined timestamp) Values: 100 bytes each (50 bytes after compression) Entries: 1000000 Prefix: 0 bytes Keys per prefix: 0 RawSize: 110.6 MB (estimated) FileSize: 62.9 MB (estimated) Write rate: 0 bytes/second Read rate: 0 ops/second Compression: Snappy Compression sampling rate: 0 Memtablerep: SkipListFactory Perf Level: 1 WARNING: Optimization is disabled: benchmarks unnecessarily slow WARNING: Assertions are enabled; benchmarks unnecessarily slow ------------------------------------------------ Initializing RocksDB Options from the specified file Initializing RocksDB Options from command-line flags Integrated BlobDB: blob cache disabled DB path: [/tmp/rocksdbtest-226125/dbbench] fillrandom : 51.271 micros/op 19504 ops/sec 51.271 seconds 1000000 operations; 2.2 MB/s waitforcompaction(/tmp/rocksdbtest-226125/dbbench): started waitforcompaction(/tmp/rocksdbtest-226125/dbbench): finished with status (OK) DB path: [/tmp/rocksdbtest-226125/dbbench] readrandom : 39.264 micros/op 25468 ops/sec 39.264 seconds 1000000 operations; 1.8 MB/s (632921 of 1000000 found) ``` Reviewed By: ajkr Differential Revision: D48524667 Pulled By: jaykorean fbshipit-source-id: 1052a15b2ed79a35165ec4d9998d0454b2552ef4 |
|
|
Changyu Bi
|
c2aad555c3 |
Add `CompressionOptions::checksum` for enabling ZSTD checksum (#11666)
Summary:
Optionally enable zstd checksum flag (
|
|
|
Peter Dillinger
|
ef6f025563 |
Placeholder for AutoHyperClockCache, more (#11692)
Summary:
* The plan is for AutoHyperClockCache to be selected when HyperClockCacheOptions::estimated_entry_charge == 0, and in that case to use a new configuration option min_avg_entry_charge for determining an extreme case maximum size for the hash table. For the placeholder, a hack is in place in HyperClockCacheOptions::MakeSharedCache() to make the unit tests happy despite the new options not really making sense with the current implementation.
* Mostly updating and refactoring tests to test both the current HCC (internal name FixedHyperClockCache) and a placeholder for the new version (internal name AutoHyperClockCache).
* Simplify some existing tests not to depend directly on cache type.
* Type-parameterize the shard-level unit tests, which unfortunately requires more syntax like `this->` in places for disambiguation.
* Added means of choosing auto_hyper_clock_cache to cache_bench, db_bench, and db_stress, including add to crash test.
* Add another templated class BaseHyperClockCache to reduce future copy-paste
* Added ReportProblems support to cache_bench
* Added a DEBUG-level diagnostic to ReportProblems for the variance in load factor throughout the table, which will become more of a concern with linear hashing to be used in the Auto implementation. Example with current Fixed HCC:
```
2023/08/10-13:41:41.602450 6ac36 [DEBUG] [che/clock_cache.cc:1507] Slot occupancy stats: Overall 49% (129008/262144), Min/Max/Window = 39%/60%/500, MaxRun{Pos/Neg} = 18/17
```
In other words, with overall occupancy of 49%, the lowest across any 500 contiguous cells is 39% and highest 60%. Longest run of occupied is 18 and longest run of unoccupied is 17. This seems consistent with random samples from a uniform distribution.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11692
Test Plan: Shouldn't be any meaningful changes yet to production code or to what is tested, but there is temporary redundancy in testing until the new implementation is plugged in.
Reviewed By: jowlyzhang
Differential Revision: D48247413
Pulled By: pdillinger
fbshipit-source-id: 11541f996d97af403c2e43c92fb67ff22dd0b5da
|
|
|
Peter Dillinger
|
99daea3481 |
Prepare tests for new HCC naming (#11676)
Summary: I'm anticipating using the public name HyperClockCache for both the current version with a fixed-size table and the upcoming version with an automatically growing table. However, for simplicity of testing them as substantially distinct implementations, I want to give them distinct internal names, like FixedHyperClockCache and AutoHyperClockCache. This change anticipates that by renaming to FixedHyperClockCache and assuming for now that all the unit tests run on HCC will run and behave similarly for the automatic HCC. Obviously updates will need to be made, but I'm trying to avoid uninteresting find & replace updates in what will be a large and engineering-heavy PR for AutoHCC Pull Request resolved: https://github.com/facebook/rocksdb/pull/11676 Test Plan: no behavior change intended, except logging will now use the name FixedHyperClockCache Reviewed By: ajkr Differential Revision: D48103165 Pulled By: pdillinger fbshipit-source-id: a33f1901488fea102164c2318e2f2b156aaba736 |
|
|
anand76
|
fcc358baf2 |
Integrate CacheReservationManager with compressed secondary cache (#11449)
Summary: This draft PR implements charging of reserved memory, for write buffers, table readers, and other purposes, proportionally to the block cache and the compressed secondary cache. The basic flow of memory reservation is maintained - clients use ```CacheReservationManager``` to request reservations, and ```CacheReservationManager``` inserts placeholder entries, i.e null value and non-zero charge, into the block cache. The ```CacheWithSecondaryAdapter``` wrapper uses its own instance of ```CacheReservationManager``` to keep track of reservations charged to the secondary cache, while the placeholder entries are inserted into the primary block cache. The design is as follows. When ```CacheWithSecondaryAdapter``` is constructed with the ```distribute_cache_res``` parameter set to true, it manages the entire memory budget across the primary and secondary cache. The secondary cache is assumed to be in memory, such as the ```CompressedSecondaryCache```. When a placeholder entry is inserted by a CacheReservationManager instance to reserve memory, the ```CacheWithSecondaryAdapter```ensures that the reservation is distributed proportionally across the primary/secondary caches. The primary block cache is initially sized to the sum of the primary cache budget + the secondary cache budget, as follows - |--------- Primary Cache Configured Capacity -----------| |---Secondary Cache Budget----|----Primary Cache Budget-----| A ```ConcurrentCacheReservationManager``` member in the ```CacheWithSecondaryAdapter```, ```pri_cache_res_```, is used to help with tracking the distribution of memory reservations. Initially, it accounts for the entire secondary cache budget as a reservation against the primary cache. This shrinks the usable capacity of the primary cache to the budget that the user originally desired. |--Reservation for Sec Cache--|-Pri Cache Usable Capacity---| When a reservation placeholder is inserted into the adapter, it is inserted directly into the primary cache. This means the entire charge of the placeholder is counted against the primary cache. To compensate and count a portion of it against the secondary cache, the secondary cache ```Deflate()``` method is called to shrink it. Since the ```Deflate()``` causes the secondary actual usage to shrink, it is reflected here by releasing an equal amount from the ```pri_cache_res_``` reservation. For example, if the pri/sec ratio is 50/50, this would be the state after placeholder insertion - |-Reservation for Sec Cache-|-Pri Cache Usable Capacity-|-R-| Likewise, when the user inserted placeholder is released, the secondary cache ```Inflate()``` method is called to grow it, and the ```pri_cache_res_``` reservation is increased by an equal amount. Other alternatives - 1. Another way of implementing this would have been to simply split the user reservation in ```CacheWithSecondaryAdapter``` into primary and secondary components. However, this would require allocating a structure to track the associated secondary cache reservation, which adds some complexity and overhead. 2. Yet another option is to implement the splitting directly in ```CacheReservationManager```. However, there are multiple instances of ```CacheReservationManager``` in a DB instance, making it complicated to keep track of them. The PR contains the following changes - 1. A new cache allocator, ```NewTieredVolatileCache()```, is defined for allocating a tiered primary block cache and compressed secondary cache. This internally allocates an instance of ```CacheWithSecondaryAdapter```. 3. New interfaces, ```Deflate()``` and ```Inflate()```, are added to the ```SecondaryCache``` interface. The default implementaion returns ```NotSupported``` with overrides in ```CompressedSecondaryCache```. 4. The ```CompressedSecondaryCache``` uses a ```ConcurrentCacheReservationManager``` instance to manage reservations done using ```Inflate()/Deflate()```. 5. The ```CacheWithSecondaryAdapter``` optionally distributes memory reservations across the primary and secondary caches. The primary cache is sized to the total memory budget (primary + secondary), and the capacity allocated to secondary cache is "reserved" against the primary cache. For any subsequent reservations, the primary cache pre-reserved capacity is adjusted. Benchmarks - Baseline ``` time ~/rocksdb_anand76/db_bench --db=/dev/shm/comp_cache_res/base --use_existing_db=true --benchmarks="readseq,readwhilewriting" --key_size=32 --value_size=1024 --num=20000000 --threads=32 --bloom_bits=10 --cache_size=30000000000 --use_compressed_secondary_cache=true --compressed_secondary_cache_size=5000000000 --duration=300 --cost_write_buffer_to_cache=true ``` ``` readseq : 3.301 micros/op 9694317 ops/sec 66.018 seconds 640000000 operations; 9763.0 MB/s readwhilewriting : 22.921 micros/op 1396058 ops/sec 300.021 seconds 418846968 operations; 1405.9 MB/s (13068999 of 13068999 found) real 6m31.052s user 152m5.660s sys 26m18.738s ``` With TieredVolatileCache ``` time ~/rocksdb_anand76/db_bench --db=/dev/shm/comp_cache_res/base --use_existing_db=true --benchmarks="readseq,readwhilewriting" --key_size=32 --value_size=1024 --num=20000000 --threads=32 --bloom_bits=10 --cache_size=30000000000 --use_compressed_secondary_cache=true --compressed_secondary_cache_size=5000000000 --duration=300 --cost_write_buffer_to_cache=true --use_tiered_volatile_cache=true ``` ``` readseq : 4.064 micros/op 7873915 ops/sec 81.281 seconds 640000000 operations; 7929.7 MB/s readwhilewriting : 20.944 micros/op 1527827 ops/sec 300.020 seconds 458378968 operations; 1538.6 MB/s (14296999 of 14296999 found) real 6m42.743s user 157m58.972s sys 33m16.671 ``` ``` readseq : 3.484 micros/op 9184967 ops/sec 69.679 seconds 640000000 operations; 9250.0 MB/s readwhilewriting : 21.261 micros/op 1505035 ops/sec 300.024 seconds 451545968 operations; 1515.7 MB/s (14101999 of 14101999 found) real 6m31.469s user 155m16.570s sys 27m47.834s ``` ToDo - 1. Add to db_stress Pull Request resolved: https://github.com/facebook/rocksdb/pull/11449 Reviewed By: pdillinger Differential Revision: D46197388 Pulled By: anand1976 fbshipit-source-id: 42d16f0254df683db4929db20d06ff26030e90df |
|
|
Peter Dillinger
|
206fdea3d9 |
Change internal headers with duplicate names (#11408)
Summary: In IDE navigation I find it annoying that there are two statistics.h files (etc.) and often land on the wrong one. Here I migrate several headers to use the blah.h <- blah_impl.h <- blah.cc idiom. Although clang-format wants "blah.h" to be the top include for "blah.cc", I think overall this is an improvement. No public API changes. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11408 Test Plan: existing tests Reviewed By: ltamasi Differential Revision: D45456696 Pulled By: pdillinger fbshipit-source-id: 809d931253f3272c908cf5facf7e1d32fc507373 |
|
|
Peter Dillinger
|
f4a02f2c52 |
Add hash_seed to Caches (#11391)
Summary: See motivation and description in new ShardedCacheOptions::hash_seed option. Updated db_bench so that its seed param is used for the cache hash seed. Made its code more safe to ensure seed is set before use. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11391 Test Plan: unit tests added / updated **Performance** - no discernible difference seen running cache_bench repeatedly before & after. With lru_cache and hyper_clock_cache. Reviewed By: hx235 Differential Revision: D45557797 Pulled By: pdillinger fbshipit-source-id: 40bf4da6d66f9d41a8a0eb8e5cf4246a4aa07934 |
|
|
clundro
|
50b33ebb1b |
remove redundant move (#11418)
Summary:
when I use g++-13 to exec the `make all` command, the output throws the warnings.
```
db/compaction/compaction_job_test.cc: In member function ‘void rocksdb::CompactionJobTestBase::AddMockFile(const rocksdb::mock::KVVector&, int)’:
db/compaction/compaction_job_test.cc:376:57: error: redundant move in initialization [-Werror=redundant-move]
376 | env_, GenerateFileName(file_number), std::move(contents)));
| ~~~~~~~~~^~~~~~~~~~
db/compaction/compaction_job_test.cc:375:7: note: in expansion of macro ‘EXPECT_OK’
375 | EXPECT_OK(mock_table_factory_->CreateMockTable(
| ^~~~~~~~~
db/compaction/compaction_job_test.cc:376:57: note: remove ‘std::move’ call
376 | env_, GenerateFileName(file_number), std::move(contents)));
| ~~~~~~~~~^~~~~~~~~~
db/compaction/compaction_job_test.cc:375:7: note: in expansion of macro ‘EXPECT_OK’
375 | EXPECT_OK(mock_table_factory_->CreateMockTable(
| ^~~~~~~~~
cc1plus: all warnings being treated as errors
make: *** [Makefile:2507: db/compaction/compaction_job_test.o] Error 1
```
and I also add some `(void)unused_variable` statements because of the cmake argument `-Wunused-but-set-variable -Wunused-but-set-variable`
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11418
Reviewed By: akankshamahajan15
Differential Revision: D45528223
Pulled By: ajkr
fbshipit-source-id: fee1a77c30039a56b481de953f0a834cc788abbc
|
|
|
Changyu Bi
|
62fc15f009 |
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 |
|
|
Peter Dillinger
|
3c17930ede |
Change default block cache from 8MB to 32MB (#11350)
Summary:
... which increases default number of shards from 16 to 64. Although the default block cache size is only recommended for applications where RocksDB is not performance-critical, under stress conditions, block cache mutex contention could become a performance bottleneck. This change of default should alleviate that.
Note that reducing the size of cache shards (recommended minimum 512MB) could cause thrashing, e.g. on filter blocks, so capacity needs to increase to safely increase number of shards.
The 8MB default dates back to 2011 or earlier (
|
|
|
sdong
|
4720ba4391 |
Remove RocksDB LITE (#11147)
Summary: We haven't been actively mantaining RocksDB LITE recently and the size must have been gone up significantly. We are removing the support. Most of changes were done through following comments: unifdef -m -UROCKSDB_LITE `git grep -l ROCKSDB_LITE | egrep '[.](cc|h)'` by Peter Dillinger. Others changes were manually applied to build scripts, CircleCI manifests, ROCKSDB_LITE is used in an expression and file db_stress_test_base.cc. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11147 Test Plan: See CI Reviewed By: pdillinger Differential Revision: D42796341 fbshipit-source-id: 4920e15fc2060c2cd2221330a6d0e5e65d4b7fe2 |
|
|
Yu Zhang
|
6943ff6e50 |
Remove deprecated util functions in options_util.h (#11126)
Summary: Remove the util functions in options_util.h that have previously been marked deprecated. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11126 Test Plan: `make check` Reviewed By: ltamasi Differential Revision: D42757496 Pulled By: jowlyzhang fbshipit-source-id: 2a138a3c207d0e0e0bbb4d99548cf2cadb44bcfb |
|
|
sdong
|
2800aa069a |
Remove compressed block cache (#11117)
Summary: Compressed block cache is replaced by compressed secondary cache. Remove the feature. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11117 Test Plan: See CI passes Reviewed By: pdillinger Differential Revision: D42700164 fbshipit-source-id: 6cbb24e460da29311150865f60ecb98637f9f67d |
|
|
leipeng
|
3941c34950 |
db_bench: let -benchmark=compact respect -subcompactions (#11077)
Summary: When running `-benchmarks=compact`, `-subcompactions` does not take effect. `-subcompactions` option comment says it is for L0-L1 compactions, it is natural to extend it to CompactionRangeOptions.max_subcompactions. This PR set CompactionRangeOptions.max_subcompactions = FLAGS_subcompactions Pull Request resolved: https://github.com/facebook/rocksdb/pull/11077 Reviewed By: akankshamahajan15 Differential Revision: D42506251 Pulled By: ajkr fbshipit-source-id: f77c9a99d32ff7af59f3c452c9e16aaeb0360304 |
|
|
Peter Dillinger
|
32520df1d9 |
Remove prototype FastLRUCache (#10954)
Summary: This was just a stepping stone to what eventually became HyperClockCache, and is now just more code to maintain. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10954 Test Plan: tests updated Reviewed By: akankshamahajan15 Differential Revision: D41310123 Pulled By: pdillinger fbshipit-source-id: 618ee148a1a0a29ee756ba8fe28359617b7cd67c |
|
|
anand76
|
aafe7bd376 |
Add multireadwhilewriting benchmark to db_bench (#10919)
Summary: Add the new benchmark Pull Request resolved: https://github.com/facebook/rocksdb/pull/10919 Reviewed By: akankshamahajan15 Differential Revision: D41017025 Pulled By: anand1976 fbshipit-source-id: 5220815d66de1f689b7f09d9c5266cebf4e345d1 |
|
|
sdong
|
48fe921754 |
Run clang format against files under tools/ and db_stress_tool/ (#10868)
Summary: Some lines of .h and .cc files are not properly fomatted. Clear them up with clang format. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10868 Test Plan: Watch existing CI to pass Reviewed By: ajkr Differential Revision: D40683485 fbshipit-source-id: 491fbb78b2cdcb948164f306829909ad816d5d0b |
|
|
Jay Zhuang
|
8124bc3526 |
Enable preclude_last_level_data_seconds in stress test (#10824)
Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/10824 Reviewed By: siying Differential Revision: D40390535 Pulled By: jay-zhuang fbshipit-source-id: 700803a1aff8a1e77c038740d87931577e79bcf6 |
|
|
Peter Dillinger
|
0f91c72adc |
Call experimental new clock cache HyperClockCache (#10684)
Summary: This change establishes a distinctive name for the experimental new lock-free clock cache (originally developed by guidotag and revamped in PR https://github.com/facebook/rocksdb/issues/10626). A few reasons: * We want to make it clear that this is a fundamentally different implementation vs. the old clock cache, to avoid people saying "I already tried clock cache." * We want to highlight the key feature: it's fast (especially under parallel load) * Because it requires an estimated charge per entry, it is not drop-in API compatible with old clock cache. This estimate might always be required for highest performance, and giving it a distinct name should reduce confusion about the distinct API requirements. * We might develop a variant requiring the same estimate parameter but with LRU eviction. In that case, using the name HyperLRUCache should make things more clear. (FastLRUCache is just a prototype that might soon be removed.) Some API detail: * To reduce copy-pasting parameter lists, etc. as in LRUCache construction, I have a `MakeSharedCache()` function on `HyperClockCacheOptions` instead of `NewHyperClockCache()`. * Changes -cache_type=clock_cache to -cache_type=hyper_clock_cache for applicable tools. I think this is more consistent / sustainable for reasons already stated. For performance tests see https://github.com/facebook/rocksdb/pull/10626 Pull Request resolved: https://github.com/facebook/rocksdb/pull/10684 Test Plan: no interesting functional changes; tests updated Reviewed By: anand1976 Differential Revision: D39547800 Pulled By: pdillinger fbshipit-source-id: 5c0fe1b5cf3cb680ab369b928c8569682b9795bf |
|
|
Peter Dillinger
|
5724348689 |
Revamp, optimize new experimental clock cache (#10626)
Summary:
* Consolidates most metadata into a single word per slot so that more
can be accomplished with a single atomic update. In the common case,
Lookup was previously about 4 atomic updates, now just 1 atomic update.
Common case Release was previously 1 atomic read + 1 atomic update,
now just 1 atomic update.
* Eliminate spins / waits / yields, which likely threaten some "lock free"
benefits. Compare-exchange loops are only used in explicit Erase, and
strict_capacity_limit=true Insert. Eviction uses opportunistic compare-
exchange.
* Relaxes some aggressiveness and guarantees. For example,
* Duplicate Inserts will sometimes go undetected and the shadow duplicate
will age out with eviction.
* In many cases, the older Inserted value for a given cache key will be kept
(i.e. Insert does not support overwrite).
* Entries explicitly erased (rather than evicted) might not be freed
immediately in some rare cases.
* With strict_capacity_limit=false, capacity limit is not tracked/enforced as
precisely as LRUCache, but is self-correcting and should only deviate by a
very small number of extra or fewer entries.
* Use smaller "computed default" number of cache shards in many cases,
because benefits to larger usage tracking / eviction pools outweigh the small
cost of more lock-free atomic contention. The improvement in CPU and I/O
is dramatic in some limit-memory cases.
* Even without the sharding change, the eviction algorithm is likely more
effective than LRU overall because it's more stateful, even though the
"hot path" state tracking for it is essentially free with ref counting. It
is like a generalized CLOCK with aging (see code comments). I don't have
performance numbers showing a specific improvement, but in theory, for a
Poisson access pattern to each block, keeping some state allows better
estimation of time to next access (Poisson interval) than strict LRU. The
bounded randomness in CLOCK can also reduce "cliff" effect for repeated
range scans approaching and exceeding cache size.
## Hot path algorithm comparison
Rough descriptions, focusing on number and kind of atomic operations:
* Old `Lookup()` (2-5 atomic updates per probe):
```
Loop:
Increment internal ref count at slot
If possible hit:
Check flags atomic (and non-atomic fields)
If cache hit:
Three distinct updates to 'flags' atomic
Increment refs for internal-to-external
Return
Decrement internal ref count
while atomic read 'displacements' > 0
```
* New `Lookup()` (1-2 atomic updates per probe):
```
Loop:
Increment acquire counter in meta word (optimistic)
If visible entry (already read meta word):
If match (read non-atomic fields):
Return
Else:
Decrement acquire counter in meta word
Else if invisible entry (rare, already read meta word):
Decrement acquire counter in meta word
while atomic read 'displacements' > 0
```
* Old `Release()` (1 atomic update, conditional on atomic read, rarely more):
```
Read atomic ref count
If last reference and invisible (rare):
Use CAS etc. to remove
Return
Else:
Decrement ref count
```
* New `Release()` (1 unconditional atomic update, rarely more):
```
Increment release counter in meta word
If last reference and invisible (rare):
Use CAS etc. to remove
Return
```
## Performance test setup
Build DB with
```
TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=fillrandom -num=30000000 -disable_wal=1 -bloom_bits=16
```
Test with
```
TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=readrandom -readonly -num=30000000 -bloom_bits=16 -cache_index_and_filter_blocks=1 -cache_size=${CACHE_MB}000000 -duration 60 -threads=$THREADS -statistics
```
Numbers on a single socket Skylake Xeon system with 48 hardware threads, DEBUG_LEVEL=0 PORTABLE=0. Very similar story on a dual socket system with 80 hardware threads. Using (every 2nd) Fibonacci MB cache sizes to sample the territory between powers of two. Configurations:
base: LRUCache before this change, but with db_bench change to default cache_numshardbits=-1 (instead of fixed at 6)
folly: LRUCache before this change, with folly enabled (distributed mutex) but on an old compiler (sorry)
gt_clock: experimental ClockCache before this change
new_clock: experimental ClockCache with this change
## Performance test results
First test "hot path" read performance, with block cache large enough for whole DB:
4181MB 1thread base -> kops/s: 47.761
4181MB 1thread folly -> kops/s: 45.877
4181MB 1thread gt_clock -> kops/s: 51.092
4181MB 1thread new_clock -> kops/s: 53.944
4181MB 16thread base -> kops/s: 284.567
4181MB 16thread folly -> kops/s: 249.015
4181MB 16thread gt_clock -> kops/s: 743.762
4181MB 16thread new_clock -> kops/s: 861.821
4181MB 24thread base -> kops/s: 303.415
4181MB 24thread folly -> kops/s: 266.548
4181MB 24thread gt_clock -> kops/s: 975.706
4181MB 24thread new_clock -> kops/s: 1205.64 (~= 24 * 53.944)
4181MB 32thread base -> kops/s: 311.251
4181MB 32thread folly -> kops/s: 274.952
4181MB 32thread gt_clock -> kops/s: 1045.98
4181MB 32thread new_clock -> kops/s: 1370.38
4181MB 48thread base -> kops/s: 310.504
4181MB 48thread folly -> kops/s: 268.322
4181MB 48thread gt_clock -> kops/s: 1195.65
4181MB 48thread new_clock -> kops/s: 1604.85 (~= 24 * 1.25 * 53.944)
4181MB 64thread base -> kops/s: 307.839
4181MB 64thread folly -> kops/s: 272.172
4181MB 64thread gt_clock -> kops/s: 1204.47
4181MB 64thread new_clock -> kops/s: 1615.37
4181MB 128thread base -> kops/s: 310.934
4181MB 128thread folly -> kops/s: 267.468
4181MB 128thread gt_clock -> kops/s: 1188.75
4181MB 128thread new_clock -> kops/s: 1595.46
Whether we have just one thread on a quiet system or an overload of threads, the new version wins every time in thousand-ops per second, sometimes dramatically so. Mutex-based implementation quickly becomes contention-limited. New clock cache shows essentially perfect scaling up to number of physical cores (24), and then each hyperthreaded core adding about 1/4 the throughput of an additional physical core (see 48 thread case). Block cache miss rates (omitted above) are negligible across the board. With partitioned instead of full filters, the maximum speed-up vs. base is more like 2.5x rather than 5x.
Now test a large block cache with low miss ratio, but some eviction is required:
1597MB 1thread base -> kops/s: 46.603 io_bytes/op: 1584.63 miss_ratio: 0.0201066 max_rss_mb: 1589.23
1597MB 1thread folly -> kops/s: 45.079 io_bytes/op: 1530.03 miss_ratio: 0.019872 max_rss_mb: 1550.43
1597MB 1thread gt_clock -> kops/s: 48.711 io_bytes/op: 1566.63 miss_ratio: 0.0198923 max_rss_mb: 1691.4
1597MB 1thread new_clock -> kops/s: 51.531 io_bytes/op: 1589.07 miss_ratio: 0.0201969 max_rss_mb: 1583.56
1597MB 32thread base -> kops/s: 301.174 io_bytes/op: 1439.52 miss_ratio: 0.0184218 max_rss_mb: 1656.59
1597MB 32thread folly -> kops/s: 273.09 io_bytes/op: 1375.12 miss_ratio: 0.0180002 max_rss_mb: 1586.8
1597MB 32thread gt_clock -> kops/s: 904.497 io_bytes/op: 1411.29 miss_ratio: 0.0179934 max_rss_mb: 1775.89
1597MB 32thread new_clock -> kops/s: 1182.59 io_bytes/op: 1440.77 miss_ratio: 0.0185449 max_rss_mb: 1636.45
1597MB 128thread base -> kops/s: 309.91 io_bytes/op: 1438.25 miss_ratio: 0.018399 max_rss_mb: 1689.98
1597MB 128thread folly -> kops/s: 267.605 io_bytes/op: 1394.16 miss_ratio: 0.0180286 max_rss_mb: 1631.91
1597MB 128thread gt_clock -> kops/s: 691.518 io_bytes/op: 9056.73 miss_ratio: 0.0186572 max_rss_mb: 1982.26
1597MB 128thread new_clock -> kops/s: 1406.12 io_bytes/op: 1440.82 miss_ratio: 0.0185463 max_rss_mb: 1685.63
610MB 1thread base -> kops/s: 45.511 io_bytes/op: 2279.61 miss_ratio: 0.0290528 max_rss_mb: 615.137
610MB 1thread folly -> kops/s: 43.386 io_bytes/op: 2217.29 miss_ratio: 0.0289282 max_rss_mb: 600.996
610MB 1thread gt_clock -> kops/s: 46.207 io_bytes/op: 2275.51 miss_ratio: 0.0290057 max_rss_mb: 637.934
610MB 1thread new_clock -> kops/s: 48.879 io_bytes/op: 2283.1 miss_ratio: 0.0291253 max_rss_mb: 613.5
610MB 32thread base -> kops/s: 306.59 io_bytes/op: 2250 miss_ratio: 0.0288721 max_rss_mb: 683.402
610MB 32thread folly -> kops/s: 269.176 io_bytes/op: 2187.86 miss_ratio: 0.0286938 max_rss_mb: 628.742
610MB 32thread gt_clock -> kops/s: 855.097 io_bytes/op: 2279.26 miss_ratio: 0.0288009 max_rss_mb: 733.062
610MB 32thread new_clock -> kops/s: 1121.47 io_bytes/op: 2244.29 miss_ratio: 0.0289046 max_rss_mb: 666.453
610MB 128thread base -> kops/s: 305.079 io_bytes/op: 2252.43 miss_ratio: 0.0288884 max_rss_mb: 723.457
610MB 128thread folly -> kops/s: 269.583 io_bytes/op: 2204.58 miss_ratio: 0.0287001 max_rss_mb: 676.426
610MB 128thread gt_clock -> kops/s: 53.298 io_bytes/op: 8128.98 miss_ratio: 0.0292452 max_rss_mb: 956.273
610MB 128thread new_clock -> kops/s: 1301.09 io_bytes/op: 2246.04 miss_ratio: 0.0289171 max_rss_mb: 788.812
The new version is still winning every time, sometimes dramatically so, and we can tell from the maximum resident memory numbers (which contain some noise, by the way) that the new cache is not cheating on memory usage. IMPORTANT: The previous generation experimental clock cache appears to hit a serious bottleneck in the higher thread count configurations, presumably due to some of its waiting functionality. (The same bottleneck is not seen with partitioned index+filters.)
Now we consider even smaller cache sizes, with higher miss ratios, eviction work, etc.
233MB 1thread base -> kops/s: 10.557 io_bytes/op: 227040 miss_ratio: 0.0403105 max_rss_mb: 247.371
233MB 1thread folly -> kops/s: 15.348 io_bytes/op: 112007 miss_ratio: 0.0372238 max_rss_mb: 245.293
233MB 1thread gt_clock -> kops/s: 6.365 io_bytes/op: 244854 miss_ratio: 0.0413873 max_rss_mb: 259.844
233MB 1thread new_clock -> kops/s: 47.501 io_bytes/op: 2591.93 miss_ratio: 0.0330989 max_rss_mb: 242.461
233MB 32thread base -> kops/s: 96.498 io_bytes/op: 363379 miss_ratio: 0.0459966 max_rss_mb: 479.227
233MB 32thread folly -> kops/s: 109.95 io_bytes/op: 314799 miss_ratio: 0.0450032 max_rss_mb: 400.738
233MB 32thread gt_clock -> kops/s: 2.353 io_bytes/op: 385397 miss_ratio: 0.048445 max_rss_mb: 500.688
233MB 32thread new_clock -> kops/s: 1088.95 io_bytes/op: 2567.02 miss_ratio: 0.0330593 max_rss_mb: 303.402
233MB 128thread base -> kops/s: 84.302 io_bytes/op: 378020 miss_ratio: 0.0466558 max_rss_mb: 1051.84
233MB 128thread folly -> kops/s: 89.921 io_bytes/op: 338242 miss_ratio: 0.0460309 max_rss_mb: 812.785
233MB 128thread gt_clock -> kops/s: 2.588 io_bytes/op: 462833 miss_ratio: 0.0509158 max_rss_mb: 1109.94
233MB 128thread new_clock -> kops/s: 1299.26 io_bytes/op: 2565.94 miss_ratio: 0.0330531 max_rss_mb: 361.016
89MB 1thread base -> kops/s: 0.574 io_bytes/op: 5.35977e+06 miss_ratio: 0.274427 max_rss_mb: 91.3086
89MB 1thread folly -> kops/s: 0.578 io_bytes/op: 5.16549e+06 miss_ratio: 0.27276 max_rss_mb: 96.8984
89MB 1thread gt_clock -> kops/s: 0.512 io_bytes/op: 4.13111e+06 miss_ratio: 0.242817 max_rss_mb: 119.441
89MB 1thread new_clock -> kops/s: 48.172 io_bytes/op: 2709.76 miss_ratio: 0.0346162 max_rss_mb: 100.754
89MB 32thread base -> kops/s: 5.779 io_bytes/op: 6.14192e+06 miss_ratio: 0.320399 max_rss_mb: 311.812
89MB 32thread folly -> kops/s: 5.601 io_bytes/op: 5.83838e+06 miss_ratio: 0.313123 max_rss_mb: 252.418
89MB 32thread gt_clock -> kops/s: 0.77 io_bytes/op: 3.99236e+06 miss_ratio: 0.236296 max_rss_mb: 396.422
89MB 32thread new_clock -> kops/s: 1064.97 io_bytes/op: 2687.23 miss_ratio: 0.0346134 max_rss_mb: 155.293
89MB 128thread base -> kops/s: 4.959 io_bytes/op: 6.20297e+06 miss_ratio: 0.323945 max_rss_mb: 823.43
89MB 128thread folly -> kops/s: 4.962 io_bytes/op: 5.9601e+06 miss_ratio: 0.319857 max_rss_mb: 626.824
89MB 128thread gt_clock -> kops/s: 1.009 io_bytes/op: 4.1083e+06 miss_ratio: 0.242512 max_rss_mb: 1095.32
89MB 128thread new_clock -> kops/s: 1224.39 io_bytes/op: 2688.2 miss_ratio: 0.0346207 max_rss_mb: 218.223
^ Now something interesting has happened: the new clock cache has gained a dramatic lead in the single-threaded case, and this is because the cache is so small, and full filters are so big, that dividing the cache into 64 shards leads to significant (random) imbalances in cache shards and excessive churn in imbalanced shards. This new clock cache only uses two shards for this configuration, and that helps to ensure that entries are part of a sufficiently big pool that their eviction order resembles the single-shard order. (This effect is not seen with partitioned index+filters.)
Even smaller cache size:
34MB 1thread base -> kops/s: 0.198 io_bytes/op: 1.65342e+07 miss_ratio: 0.939466 max_rss_mb: 48.6914
34MB 1thread folly -> kops/s: 0.201 io_bytes/op: 1.63416e+07 miss_ratio: 0.939081 max_rss_mb: 45.3281
34MB 1thread gt_clock -> kops/s: 0.448 io_bytes/op: 4.43957e+06 miss_ratio: 0.266749 max_rss_mb: 100.523
34MB 1thread new_clock -> kops/s: 1.055 io_bytes/op: 1.85439e+06 miss_ratio: 0.107512 max_rss_mb: 75.3125
34MB 32thread base -> kops/s: 3.346 io_bytes/op: 1.64852e+07 miss_ratio: 0.93596 max_rss_mb: 180.48
34MB 32thread folly -> kops/s: 3.431 io_bytes/op: 1.62857e+07 miss_ratio: 0.935693 max_rss_mb: 137.531
34MB 32thread gt_clock -> kops/s: 1.47 io_bytes/op: 4.89704e+06 miss_ratio: 0.295081 max_rss_mb: 392.465
34MB 32thread new_clock -> kops/s: 8.19 io_bytes/op: 3.70456e+06 miss_ratio: 0.20826 max_rss_mb: 519.793
34MB 128thread base -> kops/s: 2.293 io_bytes/op: 1.64351e+07 miss_ratio: 0.931866 max_rss_mb: 449.484
34MB 128thread folly -> kops/s: 2.34 io_bytes/op: 1.6219e+07 miss_ratio: 0.932023 max_rss_mb: 396.457
34MB 128thread gt_clock -> kops/s: 1.798 io_bytes/op: 5.4241e+06 miss_ratio: 0.324881 max_rss_mb: 1104.41
34MB 128thread new_clock -> kops/s: 10.519 io_bytes/op: 2.39354e+06 miss_ratio: 0.136147 max_rss_mb: 1050.52
As the miss ratio gets higher (say, above 10%), the CPU time spent in eviction starts to erode the advantage of using fewer shards (13% miss rate much lower than 94%). LRU's O(1) eviction time can eventually pay off when there's enough block cache churn:
13MB 1thread base -> kops/s: 0.195 io_bytes/op: 1.65732e+07 miss_ratio: 0.946604 max_rss_mb: 45.6328
13MB 1thread folly -> kops/s: 0.197 io_bytes/op: 1.63793e+07 miss_ratio: 0.94661 max_rss_mb: 33.8633
13MB 1thread gt_clock -> kops/s: 0.519 io_bytes/op: 4.43316e+06 miss_ratio: 0.269379 max_rss_mb: 100.684
13MB 1thread new_clock -> kops/s: 0.176 io_bytes/op: 1.54148e+07 miss_ratio: 0.91545 max_rss_mb: 66.2383
13MB 32thread base -> kops/s: 3.266 io_bytes/op: 1.65544e+07 miss_ratio: 0.943386 max_rss_mb: 132.492
13MB 32thread folly -> kops/s: 3.396 io_bytes/op: 1.63142e+07 miss_ratio: 0.943243 max_rss_mb: 101.863
13MB 32thread gt_clock -> kops/s: 2.758 io_bytes/op: 5.13714e+06 miss_ratio: 0.310652 max_rss_mb: 396.121
13MB 32thread new_clock -> kops/s: 3.11 io_bytes/op: 1.23419e+07 miss_ratio: 0.708425 max_rss_mb: 321.758
13MB 128thread base -> kops/s: 2.31 io_bytes/op: 1.64823e+07 miss_ratio: 0.939543 max_rss_mb: 425.539
13MB 128thread folly -> kops/s: 2.339 io_bytes/op: 1.6242e+07 miss_ratio: 0.939966 max_rss_mb: 346.098
13MB 128thread gt_clock -> kops/s: 3.223 io_bytes/op: 5.76928e+06 miss_ratio: 0.345899 max_rss_mb: 1087.77
13MB 128thread new_clock -> kops/s: 2.984 io_bytes/op: 1.05341e+07 miss_ratio: 0.606198 max_rss_mb: 898.27
gt_clock is clearly blowing way past its memory budget for lower miss rates and best throughput. new_clock also seems to be exceeding budgets, and this warrants more investigation but is not the use case we are targeting with the new cache. With partitioned index+filter, the miss ratio is much better, and although still high enough that the eviction CPU time is definitely offsetting mutex contention:
13MB 1thread base -> kops/s: 16.326 io_bytes/op: 23743.9 miss_ratio: 0.205362 max_rss_mb: 65.2852
13MB 1thread folly -> kops/s: 15.574 io_bytes/op: 19415 miss_ratio: 0.184157 max_rss_mb: 56.3516
13MB 1thread gt_clock -> kops/s: 14.459 io_bytes/op: 22873 miss_ratio: 0.198355 max_rss_mb: 63.9688
13MB 1thread new_clock -> kops/s: 16.34 io_bytes/op: 24386.5 miss_ratio: 0.210512 max_rss_mb: 61.707
13MB 128thread base -> kops/s: 289.786 io_bytes/op: 23710.9 miss_ratio: 0.205056 max_rss_mb: 103.57
13MB 128thread folly -> kops/s: 185.282 io_bytes/op: 19433.1 miss_ratio: 0.184275 max_rss_mb: 116.219
13MB 128thread gt_clock -> kops/s: 354.451 io_bytes/op: 23150.6 miss_ratio: 0.200495 max_rss_mb: 102.871
13MB 128thread new_clock -> kops/s: 295.359 io_bytes/op: 24626.4 miss_ratio: 0.212452 max_rss_mb: 121.109
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10626
Test Plan: updated unit tests, stress/crash test runs including with TSAN, ASAN, UBSAN
Reviewed By: anand1976
Differential Revision: D39368406
Pulled By: pdillinger
fbshipit-source-id: 5afc44da4c656f8f751b44552bbf27bd3ca6fef9
|
|
|
Levi Tamasi
|
7dad485278 |
Support JemallocNodumpAllocator for the block/blob cache in db_bench (#10685)
Summary: The patch makes it possible to use the `JemallocNodumpAllocator` with the block/blob caches in `db_bench`. In addition to its stated purpose of excluding cache contents from core dumps, `JemallocNodumpAllocator` also uses a dedicated arena and jemalloc tcaches for cache allocations, which can reduce fragmentation and thus memory usage. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10685 Reviewed By: riversand963 Differential Revision: D39552261 Pulled By: ltamasi fbshipit-source-id: b5c58eab6b7c1baa9a307d9f1248df1d7a77d2b5 |
|
|
Akanksha Mahajan
|
7a9ecdac3c |
Add auto prefetching parameters to db_bench and db_stress (#10632)
Summary: Same as title Pull Request resolved: https://github.com/facebook/rocksdb/pull/10632 Test Plan: make crash_test -j32 Reviewed By: anand1976 Differential Revision: D39241479 Pulled By: akankshamahajan15 fbshipit-source-id: 5db5b0c007da786bacc1b30d8926d36d6d029b87 |
|
|
Levi Tamasi
|
228f2c5bf5 |
Adjust the blob cache printout in db_bench/db_stress (#10614)
Summary: Currently, `db_bench` and `db_stress` print the blob cache options even if a shared block/blob cache is configured, i.e. when they are not actually in effect. The patch changes this so they are only printed when a separate blob cache is used. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10614 Test Plan: Tested manually using `db_bench` and `db_stress`. Reviewed By: akankshamahajan15 Differential Revision: D39144603 Pulled By: ltamasi fbshipit-source-id: f714304c5d46186f8514746c27ee6f52aa3e4af8 |
|
|
Changyu Bi
|
7b9e970042 |
Optionally issue `DeleteRange` in `*whilewriting` benchmarks (#10552)
Summary: Optionally issue DeleteRange in `*whilewriting` benchmarks. This happens in `BGWriter` and uses similar logic as in `DoWrite` to issue DeleteRange operations. I added this when I was benchmarking https://github.com/facebook/rocksdb/issues/10547, but this should be an independent PR. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10552 Test Plan: ran some benchmarks with various delete range options, e.g. `./db_bench --benchmarks=readwhilewriting --writes_per_range_tombstone=100 --writes=200000 --reads=1000000 --disable_auto_compactions --max_num_range_tombstones=10000` Reviewed By: ajkr Differential Revision: D38927020 Pulled By: cbi42 fbshipit-source-id: 31ee20cb8127f7173f0816ea0cc2a204ec02aad6 |
|
|
anand76
|
35cdd3e71e |
MultiGet async IO across multiple levels (#10535)
Summary: This PR exploits parallelism in MultiGet across levels. It applies only to the coroutine version of MultiGet. Previously, MultiGet file reads from SST files in the same level were parallelized. With this PR, MultiGet batches with keys distributed across multiple levels are read in parallel. This is accomplished by splitting the keys not present in a level (determined by bloom filtering) into a separate batch, and processing the new batch in parallel with the original batch. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10535 Test Plan: 1. Ensure existing MultiGet unit tests pass, updating them as necessary 2. New unit tests - TODO 3. Run stress test - TODO No noticeable regression (<1%) without async IO - Without PR: `multireadrandom : 7.261 micros/op 1101724 ops/sec 60.007 seconds 66110936 operations; 571.6 MB/s (8168992 of 8168992 found)` With PR: `multireadrandom : 7.305 micros/op 1095167 ops/sec 60.007 seconds 65717936 operations; 568.2 MB/s (8271992 of 8271992 found)` For a fully cached DB, but with async IO option on, no regression observed (<1%) - Without PR: `multireadrandom : 5.201 micros/op 1538027 ops/sec 60.005 seconds 92288936 operations; 797.9 MB/s (11540992 of 11540992 found) ` With PR: `multireadrandom : 5.249 micros/op 1524097 ops/sec 60.005 seconds 91452936 operations; 790.7 MB/s (11649992 of 11649992 found) ` Reviewed By: akankshamahajan15 Differential Revision: D38774009 Pulled By: anand1976 fbshipit-source-id: c955e259749f1c091590ade73105b3ee46cd0007 |
|
|
Akanksha Mahajan
|
5956ef0089 |
Add initial_auto_readahead_size and max_auto_readahead_size to db_bench (#10539)
Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/10539 Reviewed By: anand1976 Differential Revision: D38837111 Pulled By: akankshamahajan15 fbshipit-source-id: eb845c6e15a3c823ff6113395817388ff15a20b1 |
|
|
Gang Liao
|
275cd80cdb |
Add a blob-specific cache priority (#10461)
Summary: RocksDB's `Cache` abstraction currently supports two priority levels for items: high (used for frequently accessed/highly valuable SST metablocks like index/filter blocks) and low (used for SST data blocks). Blobs are typically lower-value targets for caching than data blocks, since 1) with BlobDB, data blocks containing blob references conceptually form an index structure which has to be consulted before we can read the blob value, and 2) cached blobs represent only a single key-value, while cached data blocks generally contain multiple KVs. Since we would like to make it possible to use the same backing cache for the block cache and the blob cache, it would make sense to add a new, lower-than-low cache priority level (bottom level) for blobs so data blocks are prioritized over them. This task is a part of https://github.com/facebook/rocksdb/issues/10156 Pull Request resolved: https://github.com/facebook/rocksdb/pull/10461 Reviewed By: siying Differential Revision: D38672823 Pulled By: ltamasi fbshipit-source-id: 90cf7362036563d79891f47be2cc24b827482743 |
|
|
Changyu Bi
|
fd165c869d |
Add memtable per key-value checksum (#10281)
Summary: Append per key-value checksum to internal key. These checksums are verified on read paths including Get, Iterator and during Flush. Get and Iterator will return `Corruption` status if there is a checksum verification failure. Flush will make DB become read-only upon memtable entry checksum verification failure. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10281 Test Plan: - Added new unit test cases: `make check` - Benchmark on memtable insert ``` TEST_TMPDIR=/dev/shm/memtable_write ./db_bench -benchmarks=fillseq -disable_wal=true -max_write_buffer_number=100 -num=10000000 -min_write_buffer_number_to_merge=100 # avg over 10 runs Baseline: 1166936 ops/sec memtable 2 bytes kv checksum : 1.11674e+06 ops/sec (-4%) memtable 2 bytes kv checksum + write batch 8 bytes kv checksum: 1.08579e+06 ops/sec (-6.95%) write batch 8 bytes kv checksum: 1.17979e+06 ops/sec (+1.1%) ``` - Benchmark on only memtable read: ops/sec dropped 31% for `readseq` due to time spend on verifying checksum. ops/sec for `readrandom` dropped ~6.8%. ``` # Readseq sudo TEST_TMPDIR=/dev/shm/memtable_read ./db_bench -benchmarks=fillseq,readseq"[-X20]" -disable_wal=true -max_write_buffer_number=100 -num=10000000 -min_write_buffer_number_to_merge=100 readseq [AVG 20 runs] : 7432840 (± 212005) ops/sec; 822.3 (± 23.5) MB/sec readseq [MEDIAN 20 runs] : 7573878 ops/sec; 837.9 MB/sec With -memtable_protection_bytes_per_key=2: readseq [AVG 20 runs] : 5134607 (± 119596) ops/sec; 568.0 (± 13.2) MB/sec readseq [MEDIAN 20 runs] : 5232946 ops/sec; 578.9 MB/sec # Readrandom sudo TEST_TMPDIR=/dev/shm/memtable_read ./db_bench -benchmarks=fillrandom,readrandom"[-X10]" -disable_wal=true -max_write_buffer_number=100 -num=1000000 -min_write_buffer_number_to_merge=100 readrandom [AVG 10 runs] : 140236 (± 3938) ops/sec; 9.8 (± 0.3) MB/sec readrandom [MEDIAN 10 runs] : 140545 ops/sec; 9.8 MB/sec With -memtable_protection_bytes_per_key=2: readrandom [AVG 10 runs] : 130632 (± 2738) ops/sec; 9.1 (± 0.2) MB/sec readrandom [MEDIAN 10 runs] : 130341 ops/sec; 9.1 MB/sec ``` - Stress test: `python3 -u tools/db_crashtest.py whitebox --duration=1800` Reviewed By: ajkr Differential Revision: D37607896 Pulled By: cbi42 fbshipit-source-id: fdaefb475629d2471780d4a5f5bf81b44ee56113 |
|
|
Jay Zhuang
|
3f763763aa |
Change `bottommost_temperture` to `last_level_temperture` (#10471)
Summary: Change tiered compaction feature from `bottommost_temperture` to `last_level_temperture`. The old option is kept for migration purpose only, which is behaving the same as `last_level_temperture` and it will be removed in the next release. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10471 Test Plan: CI Reviewed By: siying Differential Revision: D38450621 Pulled By: jay-zhuang fbshipit-source-id: cc1cdf8bad409376fec0152abc0a64fb72a91527 |
|
|
Jay Zhuang
|
1e86d424e4 |
Tiered storage stress test (#10493)
Summary: Add Tiered storage stress test and db_bench option Pull Request resolved: https://github.com/facebook/rocksdb/pull/10493 Test Plan: new crashtest: https://app.circleci.com/pipelines/github/facebook/rocksdb/16905/workflows/68c2967c-9274-434f-8506-1403cf441ead Reviewed By: ajkr Differential Revision: D38481892 Pulled By: jay-zhuang fbshipit-source-id: 217a0be4acb93d420222e6ede2a1290d9f464776 |
|
|
Changyu Bi
|
9d77bf8f7b |
Fragment memtable range tombstone in the write path (#10380)
Summary: - Right now each read fragments the memtable range tombstones https://github.com/facebook/rocksdb/issues/4808. This PR explores the idea of fragmenting memtable range tombstones in the write path and reads can just read this cached fragmented tombstone without any fragmenting cost. This PR only does the caching for immutable memtable, and does so right before a memtable is added to an immutable memtable list. The fragmentation is done without holding mutex to minimize its performance impact. - db_bench is updated to print out the number of range deletions executed if there is any. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10380 Test Plan: - CI, added asserts in various places to check whether a fragmented range tombstone list should have been constructed. - Benchmark: as this PR only optimizes immutable memtable path, the number of writes in the benchmark is chosen such an immutable memtable is created and range tombstones are in that memtable. ``` single thread: ./db_bench --benchmarks=fillrandom,readrandom --writes_per_range_tombstone=1 --max_write_buffer_number=100 --min_write_buffer_number_to_merge=100 --writes=500000 --reads=100000 --max_num_range_tombstones=100 multi_thread ./db_bench --benchmarks=fillrandom,readrandom --writes_per_range_tombstone=1 --max_write_buffer_number=100 --min_write_buffer_number_to_merge=100 --writes=15000 --reads=20000 --threads=32 --max_num_range_tombstones=100 ``` Commit |
|
|
Andrew Kryczka
|
504fe4de80 |
Avoid allocations/copies for large `GetMergeOperands()` results (#10458)
Summary: This PR avoids allocations and copies for the result of `GetMergeOperands()` when the average operand size is at least 256 bytes and the total operands size is at least 32KB. The `GetMergeOperands()` already included `PinnableSlice` but was calling `PinSelf()` (i.e., allocating and copying) for each operand. When this optimization takes effect, we instead call `PinSlice()` to skip that allocation and copy. Resources are pinned in order for the `PinnableSlice` to point to valid memory even after `GetMergeOperands()` returns. The pinned resources include a referenced `SuperVersion`, a `MergingContext`, and a `PinnedIteratorsManager`. They are bundled into a `GetMergeOperandsState`. We use `SharedCleanablePtr` to share that bundle among all `PinnableSlice`s populated by `GetMergeOperands()`. That way, the last `PinnableSlice` to be `Reset()` will cleanup the bundle, including unreferencing the `SuperVersion`. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10458 Test Plan: - new DB level test - measured benefit/regression in a number of memtable scenarios Setup command: ``` $ ./db_bench -benchmarks=mergerandom -merge_operator=StringAppendOperator -num=$num -writes=16384 -key_size=16 -value_size=$value_sz -compression_type=none -write_buffer_size=1048576000 ``` Benchmark command: ``` ./db_bench -threads=$threads -use_existing_db=true -avoid_flush_during_recovery=true -write_buffer_size=1048576000 -benchmarks=readrandomoperands -merge_operator=StringAppendOperator -num=$num -duration=10 ``` Worst regression is when a key has many tiny operands: - Parameters: num=1 (implying 16384 operands per key), value_sz=8, threads=1 - `GetMergeOperands()` latency increases 682 micros -> 800 micros (+17%) The regression disappears into the noise (<1% difference) if we remove the `Reset()` loop and the size counting loop. The former is arguably needed regardless of this PR as the convention in `Get()` and `MultiGet()` is to `Reset()` the input `PinnableSlice`s at the start. The latter could be optimized to count the size as we accumulate operands rather than after the fact. Best improvement is when a key has large operands and high concurrency: - Parameters: num=4 (implying 4096 operands per key), value_sz=2KB, threads=32 - `GetMergeOperands()` latency decreases 11492 micros -> 437 micros (-96%). Reviewed By: cbi42 Differential Revision: D38336578 Pulled By: ajkr fbshipit-source-id: 48146d127e04cb7f2d4d2939a2b9dff3aba18258 |
|
|
Peter Dillinger
|
65036e4217 |
Revert "Add a blob-specific cache priority (#10309)" (#10434)
Summary:
This reverts commit
|
|
|
Gang Liao
|
8d178090be |
Add a blob-specific cache priority (#10309)
Summary: RocksDB's `Cache` abstraction currently supports two priority levels for items: high (used for frequently accessed/highly valuable SST metablocks like index/filter blocks) and low (used for SST data blocks). Blobs are typically lower-value targets for caching than data blocks, since 1) with BlobDB, data blocks containing blob references conceptually form an index structure which has to be consulted before we can read the blob value, and 2) cached blobs represent only a single key-value, while cached data blocks generally contain multiple KVs. Since we would like to make it possible to use the same backing cache for the block cache and the blob cache, it would make sense to add a new, lower-than-low cache priority level (bottom level) for blobs so data blocks are prioritized over them. This task is a part of https://github.com/facebook/rocksdb/issues/10156 Pull Request resolved: https://github.com/facebook/rocksdb/pull/10309 Reviewed By: ltamasi Differential Revision: D38211655 Pulled By: gangliao fbshipit-source-id: 65ef33337db4d85277cc6f9782d67c421ad71dd5 |
|
|
Yanqin Jin
|
dd759537d0 |
Print perf context for all benchmarks if enabled (#10396)
Summary: If user runs `db_bench` with `-perf_level=2` or higher, db_bench should print perf context after each of all benchmarks. Or make `-perf_level` a per-benchmark switch. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10396 Test Plan: ./db_bench -benchmarks=fillseq,readseq -perf_level=2 Reviewed By: ajkr Differential Revision: D38016324 Pulled By: riversand963 fbshipit-source-id: d83ea4abc34d40ffea394ca6abf0814bc5c0a2e0 |