Summary:
Since DynamicBloom is now only used in-memory, we're free to
change it without schema compatibility issues. The new implementation
is drawn from (with manifest permission)
303542a767/bloom_simulation_tests/foo.cc (L613)
This has several speed advantages over the prior implementation:
* Uses fastrange instead of %
* Minimum logic to determine first (and all) probed memory addresses
* (Major) Two probes per 64-bit memory fetch/write.
* Very fast and effective (murmur-like) hash expansion/re-mixing. (At
least on recent CPUs, integer multiplication is very cheap.)
While a Bloom filter with 512-bit cache locality has about a 1.15x FP
rate penalty (e.g. 0.84% to 0.97%), further restricting to two probes
per 64 bits incurs an additional 1.12x FP rate penalty (e.g. 0.97% to
1.09%). Nevertheless, the unit tests show no "mediocre" FP rate samples,
unlike the old implementation with more erratic FP rates.
Especially for the memtable, we expect speed to outweigh somewhat higher
FP rates. For example, a negative table query would have to be 1000x
slower than a BF query to justify doubling BF query time to shave 10% off
FP rate (working assumption around 1% FP rate). While that seems likely
for SSTs, my data suggests a speed factor of roughly 50x for the memtable
(vs. BF; ~1.5% lower write throughput when enabling memtable Bloom
filter, after this change). Thus, it's probably not worth even 5% more
time in the Bloom filter to shave off 1/10th of the Bloom FP rate, or 0.1%
in absolute terms, and it's probably at least 20% slower to recoup that
much FP rate from this new implementation. Because of this, we do not see
a need for a 'locality' option that affects the MemTable Bloom filter
and have decoupled the MemTable Bloom filter from Options::bloom_locality.
Note that just 3% more memory to the Bloom filter (10.3 bits per key vs.
just 10) is able to make up for the ~12% FP rate drop in the new
implementation:
[] # Nearly "ideal" FP-wise but reasonably fast cache-local implementation
[~/wormhashing/bloom_simulation_tests] ./foo_gcc_IMPL_CACHE_WORM64_FROM32_any.out 10000000 6 10 $RANDOM 100000000
./foo_gcc_IMPL_CACHE_WORM64_FROM32_any.out time: 3.29372 sampled_fp_rate: 0.00985956 ...
[] # Close match to this new implementation
[~/wormhashing/bloom_simulation_tests] ./foo_gcc_IMPL_CACHE_MUL64_BLOCK_FROM32_any.out 10000000 6 10.3 $RANDOM 100000000
./foo_gcc_IMPL_CACHE_MUL64_BLOCK_FROM32_any.out time: 2.10072 sampled_fp_rate: 0.00985655 ...
[] # Old locality=1 implementation
[~/wormhashing/bloom_simulation_tests] ./foo_gcc_IMPL_CACHE_ROCKSDB_DYNAMIC_any.out 10000000 6 10 $RANDOM 100000000
./foo_gcc_IMPL_CACHE_ROCKSDB_DYNAMIC_any.out time: 3.95472 sampled_fp_rate: 0.00988943 ...
Also note the dramatic speed improvement vs. alternatives.
--
Performance unit test: DynamicBloomTest.concurrent_with_perf is updated
to report more precise timing data. (Measure running time of each
thread, not just longest running thread, etc.) Results averaged over
various sizes enabled with --enable_perf and 20 runs each; old dynamic
bloom refers to locality=1, the faster of the old:
old dynamic bloom, avg add latency = 65.6468
new dynamic bloom, avg add latency = 44.3809
old dynamic bloom, avg query latency = 50.6485
new dynamic bloom, avg query latency = 43.2186
old avg parallel add latency = 41.678
new avg parallel add latency = 24.5238
old avg parallel hit latency = 14.6322
new avg parallel hit latency = 12.3939
old avg parallel miss latency = 16.7289
new avg parallel miss latency = 12.2134
Tested on a dedicated 64-bit production machine at Facebook. Significant
improvement all around.
Despite now using std::atomic<uint64_t>, quick before-and-after test on
a 32-bit machine (Intel Atom N270, released 2008) shows no regression in
performance, in some cases modest improvement.
--
Performance integration test (synthetic): with DEBUG_LEVEL=0, used
TEST_TMPDIR=/dev/shm ./db_bench --benchmarks=fillrandom,readmissing,readrandom,stats --num=2000000
and optionally with -memtable_whole_key_filtering -memtable_bloom_size_ratio=0.01
300 runs each configuration.
Write throughput change by enabling memtable bloom:
Old locality=0: -3.06%
Old locality=1: -2.37%
New: -1.50%
conclusion -> seems to substantially close the gap
Readmissing throughput change by enabling memtable bloom:
Old locality=0: +34.47%
Old locality=1: +34.80%
New: +33.25%
conclusion -> maybe a small new penalty from FP rate
Readrandom throughput change by enabling memtable bloom:
Old locality=0: +31.54%
Old locality=1: +31.13%
New: +30.60%
conclusion -> maybe also from FP rate (after memtable flush)
--
Another conclusion we can draw from this new implementation is that the
existing 32-bit hash function is not inherently crippling the Bloom
filter speed or accuracy, below about 5 million keys. For speed, the
implementation is essentially the same whether starting with 32-bits or
64-bits of hash; it just determines whether the first multiplication
after fastrange is a pseudorandom expansion or needed re-mix. Note that
this multiplication can occur while memory is fetching.
For accuracy, in a standard configuration, you need about 5 million
keys before you have about a 1.1x FP penalty due to using a
32-bit hash vs. 64-bit:
[~/wormhashing/bloom_simulation_tests] ./foo_gcc_IMPL_CACHE_MUL64_BLOCK_FROM32_any.out $((5 * 1000 * 1000 * 10)) 6 10 $RANDOM 100000000
./foo_gcc_IMPL_CACHE_MUL64_BLOCK_FROM32_any.out time: 2.52069 sampled_fp_rate: 0.0118267 ...
[~/wormhashing/bloom_simulation_tests] ./foo_gcc_IMPL_CACHE_MUL64_BLOCK_any.out $((5 * 1000 * 1000 * 10)) 6 10 $RANDOM 100000000
./foo_gcc_IMPL_CACHE_MUL64_BLOCK_any.out time: 2.43871 sampled_fp_rate: 0.0109059
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5762
Differential Revision: D17214194
Pulled By: pdillinger
fbshipit-source-id: ad9da031772e985fd6b62a0e1db8e81892520595
Summary:
DynamicBloom was being used both for memory-only and for on-disk filters, as part of the PlainTable format. To set up enhancements to the memtable Bloom filter, this splits the code into two copies and removes unused features from each copy. Adds test PlainTableDBTest.BloomSchema to ensure no accidental change to that format.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5767
Differential Revision: D17206963
Pulled By: pdillinger
fbshipit-source-id: 6cce8d55305ed0df051b4c58bdc98c8ad81d0553
Summary:
I didn't find where customized hash function is used in DynamicBloom. This can only reduce performance. Remove it.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4915
Differential Revision: D13794452
Pulled By: siying
fbshipit-source-id: e38669b11e01444d2d782da11c7decabbd851819
Summary:
util/concurrent_arena.h:
CID 1396145 (#1 of 1): Uninitialized pointer field (UNINIT_CTOR)
2. uninit_member: Non-static class member free_begin_ is not initialized in this constructor nor in any functions that it calls.
94 Shard() : allocated_and_unused_(0) {}
util/dynamic_bloom.cc:
1. Condition hash_func == NULL, taking true branch.
CID 1322821 (#1 of 1): Uninitialized pointer field (UNINIT_CTOR)
3. uninit_member: Non-static class member data_ is not initialized in this constructor nor in any functions that it calls.
47 hash_func_(hash_func == nullptr ? &BloomHash : hash_func) {}
48
util/file_reader_writer.h:
204 private:
205 AlignedBuffer buffer_;
member_not_init_in_gen_ctor: The compiler-generated constructor for this class does not initialize buffer_offset_.
206 uint64_t buffer_offset_;
CID 1418246 (#1 of 1): Uninitialized scalar field (UNINIT_CTOR)
member_not_init_in_gen_ctor: The compiler-generated constructor for this class does not initialize buffer_len_.
207 size_t buffer_len_;
208};
util/thread_local.cc:
341#endif
CID 1322795 (#1 of 1): Uninitialized scalar field (UNINIT_CTOR)
3. uninit_member: Non-static class member pthread_key_ is not initialized in this constructor nor in any functions that it calls.
342}
40struct ThreadData {
2. uninit_member: Non-static class member next is not initialized in this constructor nor in any functions that it calls.
CID 1400668 (#1 of 1): Uninitialized pointer field (UNINIT_CTOR)
4. uninit_member: Non-static class member prev is not initialized in this constructor nor in any functions that it calls.
41 explicit ThreadData(ThreadLocalPtr::StaticMeta* _inst) : entries(), inst(_inst) {}
42 std::vector<Entry> entries;
1. member_decl: Class member declaration for next.
43 ThreadData* next;
3. member_decl: Class member declaration for prev.
44 ThreadData* prev;
45 ThreadLocalPtr::StaticMeta* inst;
46};
Closes https://github.com/facebook/rocksdb/pull/3123
Differential Revision: D6233566
Pulled By: sagar0
fbshipit-source-id: aa2068790ea69787a0035c0db39d59b0c25108db
Summary:
This diff adds support for concurrent adds to the skiplist memtable
implementations. Memory allocation is made thread-safe by the addition of
a spinlock, with small per-core buffers to avoid contention. Concurrent
memtable writes are made via an additional method and don't impose a
performance overhead on the non-concurrent case, so parallelism can be
selected on a per-batch basis.
Write thread synchronization is an increasing bottleneck for higher levels
of concurrency, so this diff adds --enable_write_thread_adaptive_yield
(default off). This feature causes threads joining a write batch
group to spin for a short time (default 100 usec) using sched_yield,
rather than going to sleep on a mutex. If the timing of the yield calls
indicates that another thread has actually run during the yield then
spinning is avoided. This option improves performance for concurrent
situations even without parallel adds, although it has the potential to
increase CPU usage (and the heuristic adaptation is not yet mature).
Parallel writes are not currently compatible with
inplace updates, update callbacks, or delete filtering.
Enable it with --allow_concurrent_memtable_write (and
--enable_write_thread_adaptive_yield). Parallel memtable writes
are performance neutral when there is no actual parallelism, and in
my experiments (SSD server-class Linux and varying contention and key
sizes for fillrandom) they are always a performance win when there is
more than one thread.
Statistics are updated earlier in the write path, dropping the number
of DB mutex acquisitions from 2 to 1 for almost all cases.
This diff was motivated and inspired by Yahoo's cLSM work. It is more
conservative than cLSM: RocksDB's write batch group leader role is
preserved (along with all of the existing flush and write throttling
logic) and concurrent writers are blocked until all memtable insertions
have completed and the sequence number has been advanced, to preserve
linearizability.
My test config is "db_bench -benchmarks=fillrandom -threads=$T
-batch_size=1 -memtablerep=skip_list -value_size=100 --num=1000000/$T
-level0_slowdown_writes_trigger=9999 -level0_stop_writes_trigger=9999
-disable_auto_compactions --max_write_buffer_number=8
-max_background_flushes=8 --disable_wal --write_buffer_size=160000000
--block_size=16384 --allow_concurrent_memtable_write" on a two-socket
Xeon E5-2660 @ 2.2Ghz with lots of memory and an SSD hard drive. With 1
thread I get ~440Kops/sec. Peak performance for 1 socket (numactl
-N1) is slightly more than 1Mops/sec, at 16 threads. Peak performance
across both sockets happens at 30 threads, and is ~900Kops/sec, although
with fewer threads there is less performance loss when the system has
background work.
Test Plan:
1. concurrent stress tests for InlineSkipList and DynamicBloom
2. make clean; make check
3. make clean; DISABLE_JEMALLOC=1 make valgrind_check; valgrind db_bench
4. make clean; COMPILE_WITH_TSAN=1 make all check; db_bench
5. make clean; COMPILE_WITH_ASAN=1 make all check; db_bench
6. make clean; OPT=-DROCKSDB_LITE make check
7. verify no perf regressions when disabled
Reviewers: igor, sdong
Reviewed By: sdong
Subscribers: MarkCallaghan, IslamAbdelRahman, anthony, yhchiang, rven, sdong, guyg8, kradhakrishnan, dhruba
Differential Revision: https://reviews.facebook.net/D50589
Summary:
Introduces a new class for managing write buffer memory across column
families. We supplement ColumnFamilyOptions::write_buffer_size with
ColumnFamilyOptions::write_buffer, a shared pointer to a WriteBuffer
instance that enforces memory limits before flushing out to disk.
Test Plan: Added SharedWriteBuffer unit test to db_test.cc
Reviewers: sdong, rven, ljin, igor
Reviewed By: igor
Subscribers: tnovak, yhchiang, dhruba, xjin, MarkCallaghan, yoshinorim
Differential Revision: https://reviews.facebook.net/D22581
* Script for building the unity.cc file via Makefile
* Unity executable Makefile target for testing builds
* Source code changes to fix compilation of unity build
Summary:
Adding option to save PlainTable index and bloom filter in SST file.
If there is no bloom block and/or index block, PlainTableReader builds
new ones. Otherwise PlainTableReader just use these blocks.
Test Plan: make all check
Reviewers: sdong
Reviewed By: sdong
Subscribers: leveldb
Differential Revision: https://reviews.facebook.net/D19527
Summary:
Bloomfilter and hashskiplist's buckets_ allocated by memtable's arena
DynamicBloom: pass arena via constructor, allocate space in SetTotalBits
HashSkipListRep: allocate space of buckets_ using arena.
do not delete it in deconstructor because arena would take care of it.
Several test files are changed.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: igor, dhruba
Differential Revision: https://reviews.facebook.net/D19335
Summary:
Clean PlainTableReader's data structures:
(1) inline bloom_ (in order to do this, change DynamicBloom to allow lazy initialization)
(2) remove some variables only used when initialization from the class
(3) put variables not used in normal read code paths to the end of the class and reference prefix_extractor directly
(4) make Options a reference.
Test Plan: make all check
Reviewers: haobo, ljin
Reviewed By: ljin
Subscribers: igor, yhchiang, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D18891
Summary:
This patch changes meaning of options.bloom_locality: 0 means disable cache line optimization and any positive number means use CACHE_LINE_SIZE as block size (the previous behavior is the block size will be CACHE_LINE_SIZE*options.bloom_locality). By doing it, the divide operations inside a block can be replaced by a shift.
Performance is improved:
https://reviews.facebook.net/P471
Also, improve the basic algorithm in two ways:
(1) make sure num of blocks is an odd number
(2) rotate bytes after every probe in locality mode. Since the divider is 2^n, unless doing it, we are never able to use all the bits.
Improvements of false positive: https://reviews.facebook.net/P459
Test Plan: make all check
Reviewers: ljin, haobo
Reviewed By: haobo
Subscribers: dhruba, yhchiang, igor, leveldb
Differential Revision: https://reviews.facebook.net/D18843
Summary:
TLB page allocation errors are now logged to info logs, instead of stderr.
In order to do that, mem table rep's factory functions take a info logger now.
Test Plan: make all check
Reviewers: haobo, igor, yhchiang
Reviewed By: yhchiang
CC: leveldb, yhchiang, dhruba
Differential Revision: https://reviews.facebook.net/D18471
Summary: Add an option to allocate a piece of memory from huge page TLB. Add options to trigger it in dynamic bloom, plain table indexes andhash linked list hash table.
Test Plan: make all check
Reviewers: haobo, ljin
Reviewed By: haobo
CC: nkg-, dhruba, leveldb, igor, yhchiang
Differential Revision: https://reviews.facebook.net/D18357
Summary: this causes overflow and asan failure
Test Plan: make asan_check
Reviewers: igor
Reviewed By: igor
CC: leveldb
Differential Revision: https://reviews.facebook.net/D17301
Summary:
By constraining the probes within cache line(s), we can improve the
cache miss rate thus performance. This probably only makes sense for
in-memory workload so defaults the option to off.
Numbers and comparision can be found in wiki:
https://our.intern.facebook.com/intern/wiki/index.php/Ljin/rocksdb_perf/2014_03_17#Bloom_Filter_Study
Test Plan: benchmarked this change substantially. Will run make all check as well
Reviewers: haobo, igor, dhruba, sdong, yhchiang
Reviewed By: haobo
CC: leveldb
Differential Revision: https://reviews.facebook.net/D17133
Summary:
In latest leaf's, MayContainHash() consistently consumes 5%~7% CPU usage.
I checked the code and did an experiment with/without inlining this method.
In release mode, with `1024 * 1024 * 256` bits and `1024 * 512` entries, both call 2^30 MayContainHash() with distinctive parameters.
As the result showed, this patch reduced the running time from 9.127 sec to 7.891 sec.
Test Plan: make check
Reviewers: sdong, haobo
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15177
Summary:
Some changes to PlainTable format:
(1) support variable key length
(2) use user defined slice transformer to extract prefixes
(3) Run some test cases against PlainTable in db_test and table_test
Test Plan: test db_test
Reviewers: haobo, kailiu
CC: dhruba, igor, leveldb, nkg-
Differential Revision: https://reviews.facebook.net/D14457
Summary: as title
Test Plan: dynamic_bloom_test
Reviewers: dhruba, sdong, kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D14385
Peter Dillinger