Summary:
This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code.
A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest.
TODO:
1. Figure out how to build it in CircleCI (requires some dependencies to be installed)
2. Do some stress testing with coroutines enabled
No regression in synchronous MultiGet between this branch and main -
```
./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics
```
Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)```
Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)```
More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file.
1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) -
No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)```
Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)```
2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file -
No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)```
Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)```
3. Single thread CPU bound workload with ~2 key overlap/file -
No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)```
Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)```
4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file -
No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ```
Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968
Reviewed By: akankshamahajan15
Differential Revision: D36348563
Pulled By: anand1976
fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
Summary:
The P95 and P99 metrics are flaky, similar to DBGet ones which removed
in https://github.com/facebook/rocksdb/issues/9742 .
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9844
Test Plan: `$ ./buckifier/buckify_rocksdb.py`
Reviewed By: ajkr
Differential Revision: D35655531
Pulled By: jay-zhuang
fbshipit-source-id: c1409f0fba4e23d461a65f988c27ac5e2ae85d13
Summary:
Especially after updating to C++17, I don't see a compelling case for
*requiring* any folly components in RocksDB. I was able to purge the existing
hard dependencies, and it can be quite difficult to strip out non-trivial components
from folly for use in RocksDB. (The prospect of doing that on F14 has changed
my mind on the best approach here.)
But this change creates an optional integration where we can plug in
components from folly at compile time, starting here with F14FastMap to replace
std::unordered_map when possible (probably no public APIs for example). I have
replaced the biggest CPU users of std::unordered_map with compile-time
pluggable UnorderedMap which will use F14FastMap when USE_FOLLY is set.
USE_FOLLY is always set in the Meta-internal buck build, and a simulation of
that is in the Makefile for public CI testing. A full folly build is not needed, but
checking out the full folly repo is much simpler for getting the dependency,
and anything else we might want to optionally integrate in the future.
Some picky details:
* I don't think the distributed mutex stuff is actually used, so it was easy to remove.
* I implemented an alternative to `folly::constexpr_log2` (which is much easier
in C++17 than C++11) so that I could pull out the hard dependencies on
`ConstexprMath.h`
* I had to add noexcept move constructors/operators to some types to make
F14's complainUnlessNothrowMoveAndDestroy check happy, and I added a
macro to make that easier in some common cases.
* Updated Meta-internal buck build to use folly F14Map (always)
No updates to HISTORY.md nor INSTALL.md as this is not (yet?) considered a
production integration for open source users.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9546
Test Plan:
CircleCI tests updated so that a couple of them use folly.
Most internal unit & stress/crash tests updated to use Meta-internal latest folly.
(Note: they should probably use buck but they currently use Makefile.)
Example performance improvement: when filter partitions are pinned in cache,
they are tracked by PartitionedFilterBlockReader::filter_map_ and we can build
a test that exercises that heavily. Build DB with
```
TEST_TMPDIR=/dev/shm/rocksdb ./db_bench -benchmarks=fillrandom -num=10000000 -disable_wal=1 -write_buffer_size=30000000 -bloom_bits=16 -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=0 -partition_index_and_filters
```
and test with (simultaneous runs with & without folly, ~20 times each to see
convergence)
```
TEST_TMPDIR=/dev/shm/rocksdb ./db_bench_folly -readonly -use_existing_db -benchmarks=readrandom -num=10000000 -bloom_bits=16 -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=0 -partition_index_and_filters -duration=40 -pin_l0_filter_and_index_blocks_in_cache
```
Average ops/s no folly: 26229.2
Average ops/s with folly: 26853.3 (+2.4%)
Reviewed By: ajkr
Differential Revision: D34181736
Pulled By: pdillinger
fbshipit-source-id: ffa6ad5104c2880321d8a1aa7187e00ab0d02e94
Summary:
DBGet p95 and p99 have high variation, remove them for now.
Also increase the iteration to 3 to avoid false positive.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9742
Test Plan: Internal CI
Reviewed By: ajkr
Differential Revision: D35082820
Pulled By: jay-zhuang
fbshipit-source-id: facc1d56b94e54aa8c8852c207aae2ae4e4924b0
Summary:
After https://github.com/facebook/rocksdb/issues/9515 added a unique_ptr to Status, we see some
warnings-as-error in some internal builds like this:
```
stderr: rocksdb/src/db/compaction/compaction_job.cc:2839:7: error:
offset of on non-standard-layout type 'struct CompactionServiceResult'
[-Werror,-Winvalid-offsetof]
{offsetof(struct CompactionServiceResult, status),
^ ~~~~~~
```
I see three potential solutions to resolving this:
* Expand our use of an idiom that works around the warning (see offset_of
functions removed in this change, inspired by
https://gist.github.com/graphitemaster/494f21190bb2c63c5516) However,
this construction is invoking undefined behavior that assumes consistent
layout with no compiler-introduced indirection. A compiler incompatible
with our assumptions will likely compile the code and exhibit undefined
behavior.
* Migrate to something in place of offset, like a function mapping
CompactionServiceResult* to Status* (for the `status` field). This might
be required in the long term.
* **Selected:** Use our new C++17 dependency to use offsetof in a well-defined way
when the compiler allows it. From a comment on
https://gist.github.com/graphitemaster/494f21190bb2c63c5516:
> A final note: in C++17, offsetof is conditionally supported, which
> means that you can use it on any type (not just standard layout
> types) and the compiler will error if it can't compile it correctly.
> That appears to be the best option if you can live with C++17 and
> don't need constexpr support.
The C++17 semantics are confirmed on
https://en.cppreference.com/w/cpp/types/offsetof, so we can suppress the
warning as long as we accept that we might run into a compiler that
rejects the code, and at that point we will find a solution, such as
the more intrusive "migrate" solution above.
Although this is currently only showing in our buck build, it will
surely show up also with make and cmake, so I have updated those
configurations as well.
Also in the buck build, -Wno-expansion-to-defined does not appear to be
needed anymore (both current compiler configurations) so I
removed it.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9563
Test Plan: Tried out buck builds with both current compiler configurations
Reviewed By: riversand963
Differential Revision: D34220931
Pulled By: pdillinger
fbshipit-source-id: d39436008259bd1eaaa87c77be69fb2a5b559e1f
Summary:
49af999954
updates RocksDB buckifier script directly via fbcode. We need to make
sure that the following command run in RocksDB repo generate the same
TARGETS file.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9104
Test Plan:
```
$python buckifier/buckify_rocksdb.py
```
Verify that TARGETS file does not have uncommitted changes.
Reviewed By: jay-zhuang
Differential Revision: D32055387
Pulled By: riversand963
fbshipit-source-id: 19cf1b8145095b6df625958458189680e543e3ba
Summary:
This PR adds a ```-secondary_cache_uri``` option to the cache_bench and db_bench tools to allow the user to specify a custom secondary cache URI. The object registry is used to create an instance of the ```SecondaryCache``` object of the type specified in the URI.
The main cache_bench code is packaged into a separate library, similar to db_bench.
An example invocation of db_bench with a secondary cache URI -
```db_bench --env_uri=ws://ws.flash_sandbox.vll1_2/ -db=anand/nvm_cache_2 -use_existing_db=true -benchmarks=readrandom -num=30000000 -key_size=32 -value_size=256 -use_direct_reads=true -cache_size=67108864 -cache_index_and_filter_blocks=true -secondary_cache_uri='cachelibwrapper://filename=/home/anand76/nvm_cache/cache_file;size=2147483648;regionSize=16777216;admPolicy=random;admProbability=1.0;volatileSize=8388608;bktPower=20;lockPower=12' -partition_index_and_filters=true -duration=1800```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8312
Reviewed By: zhichao-cao
Differential Revision: D28544325
Pulled By: anand1976
fbshipit-source-id: 8f209b9af900c459dc42daa7a610d5f00176eeed
Summary:
New tests should by default be expected to be parallelizeable
and passing with ASSERT_STATUS_CHECKED. Thus, I'm changing those two
lists to exclusions rather than inclusions.
For the set of exclusions, I only listed things that currently failed
for me when attempting not to exclude, or had some other documented
reason. This marks many more tests as "parallel," which will potentially
cause some failures from self-interference, but we can address those as
they are discovered.
Also changed CircleCI ASC test to be parallelized; the easy way to do
that is to exclude building tests that don't pass ASC, which is now a
small set.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8146
Test Plan: Watch CI, etc.
Reviewed By: riversand963
Differential Revision: D27542782
Pulled By: pdillinger
fbshipit-source-id: bdd74bcd912a963ee33f3fc0d2cad2567dc7740f
Summary:
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8143
The latter assume the location of the compile root, which can break
if the build root changes. Switch to the slightly more intelligent
`include_paths`, which should provide the same functionality, but do
with independent of include root.
Reviewed By: riversand963
Differential Revision: D27535869
fbshipit-source-id: 0129e47c0ce23e08528c9139114a591c14866fa8
Summary:
If the platform is ppc64 and the libc is not GNU libc, then we exclude the range_tree from compilation.
See https://jira.percona.com/browse/PS-7559
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8070
Reviewed By: jay-zhuang
Differential Revision: D27246004
Pulled By: mrambacher
fbshipit-source-id: 59d8433242ce7ce608988341becb4f83312445f5
Summary:
TIL we have different versions of TARGETS file generated with
options passed to buckifier. Someone thought they were totally fine to
squash the file by re-running the command to generate (pretty reasonable
assumption) but the command was incorrect due to missing the extra
argument used to generate THAT TARGETS file.
This change includes in the command written in the TARGETS header the
extra argument passed to buckify (when used).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7902
Test Plan:
manual, as in the (now fixed) comments at the top of
buckify_rocksdb.py
Reviewed By: ajkr
Differential Revision: D26108317
Pulled By: pdillinger
fbshipit-source-id: 46e93dc1465e27bd18e0e0baa8eeee1b591c765d
Summary:
Prefer to use keyword args rather than positional args for Buck rules. This appears to be the only remaining instance for `custom_unittest`
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7760
Test Plan: Search for other instances of `custom_unittest` without `name`
Reviewed By: cheng-chang
Differential Revision: D25439887
Pulled By: mzlee
fbshipit-source-id: 518c541a5c01207c7b0c1f7322addf5cc4f09f92
Summary:
Buck TARGETS files are sometimes parsed with Python, and sometimes with Starlark - this TARGETS file was not Starlark compliant. In Starlark you can't have a top-level if in a TARGETS file, but you can have a ternary `a if b else c`. Therefore I converted TARGETS, and updated the generator for it.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7743
Reviewed By: pdillinger
Differential Revision: D25342587
Pulled By: ndmitchell
fbshipit-source-id: 88cbe8632071a45a3ea8675812967614c62c78d1
Summary:
We would like to build a shared library with all fbcode dependencies statically linked within.
This resulting .so should not drop any symbols definitions in the building process.
To ensure that, we use `link_whole=True` according to
https://buck.build/rule/cxx_library.html#link_whole.
Since `link_whole` is `False` by default, adding a `link_whole=False` to existing libraries won't
change any behavior.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7466
Test Plan: build a .so and test internally.
Reviewed By: pdillinger
Differential Revision: D24009780
Pulled By: riversand963
fbshipit-source-id: d18804d495da7195ed72a2040e1a5de4fd336519
Summary:
Re-add extra_compiler_flags when building unit tests for fbcode.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7437
Test Plan: Integrate with buck and run internal tests.
Reviewed By: pdillinger
Differential Revision: D23943924
Pulled By: riversand963
fbshipit-source-id: b92b7ad003e06e0860c45efc5f7f9684233d0c55
Summary:
While rocksdb can compile on both macOS and Linux with Buck, it couldn't be
compiled on Windows. The only way to compile it on Windows was with the CMake
build.
To keep the multi-platform complexity low, I've simply included all the Windows
bits in the TARGETS file, and added large #if blocks when not on Windows, the
same was done on the posix specific files.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7406
Test Plan:
On my devserver:
buck test //rocksdb/...
On Windows:
buck build mode/win //rocksdb/src:rocksdb_lib
Reviewed By: pdillinger
Differential Revision: D23874358
Pulled By: xavierd
fbshipit-source-id: 8768b5d16d7e8f44b5ca1e2483881ca4b24bffbe
Summary:
This is to fix special logic to run tests inside FB.
Buck test is broken after moving to cpp_unittest(). Move c_test back to the previous approach.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7076
Test Plan: Watch the Sandcastle run
Reviewed By: ajkr
Differential Revision: D22370096
fbshipit-source-id: 4a464d0903f2c76ae2de3a8ad373ffc9bedec64c
Summary:
Change the linking of tests/tools to be against a library rather than a list of objects. This change substantially reduces the size of the objects produced.
peterd clean repo size: 264M
Before this change, with make all: 40G
After this change, with make all: 28G
With make LIB_MODE=shared all: 7.0G
The list of TESTS was changed from being hard-coded to generated from the test sources variable. Note that there are some test sources that are not built as tests (though the set of tests is identical to the previous version).
Added OBJ_DIR option to Makefile to allow objects to be placed in an alternative location. By default, OBJ_DIR is the same as before ("./").
This change is a precursor to being able to build/run the tests/tools linked against static libraries. Additionally, it should be possible to clean up and merge some of the rules for building tests and the like if so desired.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6660
Reviewed By: riversand963
Differential Revision: D22244463
Pulled By: pdillinger
fbshipit-source-id: db9c6341d81ed62c2270374f4ede02fb9604c754
Summary:
Make RocksDB run a predefined unit test so that it can be integrated with better tools.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6926
Test Plan: Watch tests
Reviewed By: pdillinger
Differential Revision: D21866216
fbshipit-source-id: cafca82efdf0b72671be8d30b665e88a75ae6000
Summary:
RocksDB Makefile was assuming existence of 'python' command,
which is not present in CentOS 8. We avoid using 'python' if 'python3' is available.
Also added fancy logic to format-diff.sh to make clang-format-diff.py for Python2 work even with Python3 only (as some CentOS 8 FB machines come equipped)
Also, now use just 'python3' for PYTHON if not found so that an informative
"command not found" error will result rather than something weird.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6883
Test Plan: manually tried some variants, 'make check' on a fresh CentOS 8 machine without 'python' executable or Python2 but with clang-format-diff.py for Python2.
Reviewed By: gg814
Differential Revision: D21767029
Pulled By: pdillinger
fbshipit-source-id: 54761b376b140a3922407bdc462f3572f461d0e9
Summary:
... so that we have freedom to upgrade it (see https://github.com/facebook/rocksdb/issues/6808).
As a side benefit, gtest will no longer be linked into main library in
buck build.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6858
Test Plan: fb internal build & link
Reviewed By: riversand963
Differential Revision: D21652061
Pulled By: pdillinger
fbshipit-source-id: 6018104af944debde576b5beda6c134e737acedb
Summary:
In buck build with opt mode, target should not include rocksdb_test_lib.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6847
Test Plan: Watch for internal cont build.
Reviewed By: ajkr
Differential Revision: D21586803
Pulled By: riversand963
fbshipit-source-id: 76d253c18d16fac6cab86a8c3f6b471ad5b6efb3
Summary:
Before this PR, extra deps passed in from cmd line to buckifier will be parsed
and used to populate a dict. Using this dict and printing to TARGETS file will
lead to printing u'', disallowed by build tools. This PR removes the u''.
Test Plan (local dev server):
```
python buckifier/buckify_rocksdb.py '{"fake": {"extra_deps": [":test_dep", "//fake/module:mock1"], "extra_compiler_flags": ["-Os", "-DROCKSDB_LITE"]}}'
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6841
Reviewed By: siying
Differential Revision: D21538155
Pulled By: riversand963
fbshipit-source-id: 09403668a4aa1a15bad7dac229c2bc8ce8ee1349
Summary:
Add Github Action to perform some basic sanity check for PR, inclding the
following.
1) Buck TARGETS file.
On the one hand, The TARGETS file is used for internal buck, and we do not
manually update it. On the other hand, we need to run the buckifier scripts to
update TARGETS whenever new files are added, etc. With this Github Action, we
make sure that every PR does not forget this step. The GH Action uses
a Makefile target called check-buck-targets. Users can manually run `make
check-buck-targets` on local machine.
2) Code format
We use clang-format-diff.py to format our code. The GH Action in this PR makes
sure this step is not skipped. The checking script build_tools/format-diff.sh assumes that `clang-format-diff.py` is executable.
On host running GH Action, it is difficult to download `clang-format-diff.py` and make it
executable. Therefore, we modified build_tools/format-diff.sh to handle the case in which there is a non-executable clang-format-diff.py file in the top-level rocksdb repo directory.
Test Plan (Github and devserver):
Watch for Github Action result in the `Checks` tab.
On dev server
```
make check-format
make check-buck-targets
make check
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6761
Test Plan: Watch for Github Action result in the `Checks` tab.
Reviewed By: pdillinger
Differential Revision: D21260209
Pulled By: riversand963
fbshipit-source-id: c646e2f37c6faf9f0614b68aa0efc818cff96787
Summary:
Some recent PRs added new source files or modified TARGETS file manually.
During next internal release, executing the following command will revert the
manual changes.
Update buckifier so that the following command
```
python buckfier/buckify_rocksdb.py
```
does not change TARGETS file.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6726
Test Plan:
```
python buckifier/buckify_rocksdb.py
```
Reviewed By: siying
Differential Revision: D21098930
Pulled By: riversand963
fbshipit-source-id: e884f507fefef88163363c9097a460c98f1ed850
Summary:
The known bug of liburing has been fixed. Now we can re-enable liburing under Linux
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6451
Test Plan: Watch internal CI
Differential Revision: D20079009
fbshipit-source-id: 04a6f53a900ff721f9a62a188cf906771b5d68d2
Summary:
Since IO Uring feature is not stable. Remove it from buck configuration.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6415
Test Plan: See internal build pass
Differential Revision: D19892988
fbshipit-source-id: 7fc01efc2af5ed707fb8e4e4674223aeb83cd5ea
Summary:
Right now, PosixRandomAccessFile::MultiRead() executes read requests in parallel. In this PR, it leverages I/O Uring library to run it in parallel, even when page cache is enabled. This function will fall back if the kernel version doesn't support it.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5881
Test Plan: Run the unit test on a kernel version supporting it and make sure all tests pass, and run a unit test on kernel version supporting it and see it pass. Before merging, will also run stress test and see it passes.
Differential Revision: D17742266
fbshipit-source-id: e05699c925ac04fdb42379456a4e23e4ebcb803a
Summary:
Had complications with LITE build and valgrind test.
Reverts/fixes small parts of PR https://github.com/facebook/rocksdb/issues/6007
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6036
Test Plan:
make LITE=1 all check
and
ROCKSDB_VALGRIND_RUN=1 DISABLE_JEMALLOC=1 make -j24 db_bloom_filter_test && ROCKSDB_VALGRIND_RUN=1 DISABLE_JEMALLOC=1 ./db_bloom_filter_test
Differential Revision: D18512238
Pulled By: pdillinger
fbshipit-source-id: 37213cf0d309edf11c483fb4b2fb6c02c2cf2b28
Summary:
Adds an improved, replacement Bloom filter implementation (FastLocalBloom) for full and partitioned filters in the block-based table. This replacement is faster and more accurate, especially for high bits per key or millions of keys in a single filter.
Speed
The improved speed, at least on recent x86_64, comes from
* Using fastrange instead of modulo (%)
* Using our new hash function (XXH3 preview, added in a previous commit), which is much faster for large keys and only *slightly* slower on keys around 12 bytes if hashing the same size many thousands of times in a row.
* Optimizing the Bloom filter queries with AVX2 SIMD operations. (Added AVX2 to the USE_SSE=1 build.) Careful design was required to support (a) SIMD-optimized queries, (b) compatible non-SIMD code that's simple and efficient, (c) flexible choice of number of probes, and (d) essentially maximized accuracy for a cache-local Bloom filter. Probes are made eight at a time, so any number of probes up to 8 is the same speed, then up to 16, etc.
* Prefetching cache lines when building the filter. Although this optimization could be applied to the old structure as well, it seems to balance out the small added cost of accumulating 64 bit hashes for adding to the filter rather than 32 bit hashes.
Here's nominal speed data from filter_bench (200MB in filters, about 10k keys each, 10 bits filter data / key, 6 probes, avg key size 24 bytes, includes hashing time) on Skylake DE (relatively low clock speed):
$ ./filter_bench -quick -impl=2 -net_includes_hashing # New Bloom filter
Build avg ns/key: 47.7135
Mixed inside/outside queries...
Single filter net ns/op: 26.2825
Random filter net ns/op: 150.459
Average FP rate %: 0.954651
$ ./filter_bench -quick -impl=0 -net_includes_hashing # Old Bloom filter
Build avg ns/key: 47.2245
Mixed inside/outside queries...
Single filter net ns/op: 63.2978
Random filter net ns/op: 188.038
Average FP rate %: 1.13823
Similar build time but dramatically faster query times on hot data (63 ns to 26 ns), and somewhat faster on stale data (188 ns to 150 ns). Performance differences on batched and skewed query loads are between these extremes as expected.
The only other interesting thing about speed is "inside" (query key was added to filter) vs. "outside" (query key was not added to filter) query times. The non-SIMD implementations are substantially slower when most queries are "outside" vs. "inside". This goes against what one might expect or would have observed years ago, as "outside" queries only need about two probes on average, due to short-circuiting, while "inside" always have num_probes (say 6). The problem is probably the nastily unpredictable branch. The SIMD implementation has few branches (very predictable) and has pretty consistent running time regardless of query outcome.
Accuracy
The generally improved accuracy (re: Issue https://github.com/facebook/rocksdb/issues/5857) comes from a better design for probing indices
within a cache line (re: Issue https://github.com/facebook/rocksdb/issues/4120) and improved accuracy for millions of keys in a single filter from using a 64-bit hash function (XXH3p). Design details in code comments.
Accuracy data (generalizes, except old impl gets worse with millions of keys):
Memory bits per key: FP rate percent old impl -> FP rate percent new impl
6: 5.70953 -> 5.69888
8: 2.45766 -> 2.29709
10: 1.13977 -> 0.959254
12: 0.662498 -> 0.411593
16: 0.353023 -> 0.0873754
24: 0.261552 -> 0.0060971
50: 0.225453 -> ~0.00003 (less than 1 in a million queries are FP)
Fixes https://github.com/facebook/rocksdb/issues/5857
Fixes https://github.com/facebook/rocksdb/issues/4120
Unlike the old implementation, this implementation has a fixed cache line size (64 bytes). At 10 bits per key, the accuracy of this new implementation is very close to the old implementation with 128-byte cache line size. If there's sufficient demand, this implementation could be generalized.
Compatibility
Although old releases would see the new structure as corrupt filter data and read the table as if there's no filter, we've decided only to enable the new Bloom filter with new format_version=5. This provides a smooth path for automatic adoption over time, with an option for early opt-in.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6007
Test Plan: filter_bench has been used thoroughly to validate speed, accuracy, and correctness. Unit tests have been carefully updated to exercise new and old implementations, as well as the logic to select an implementation based on context (format_version).
Differential Revision: D18294749
Pulled By: pdillinger
fbshipit-source-id: d44c9db3696e4d0a17caaec47075b7755c262c5f
Summary:
include db_stress_tool in rocksdb tools lib
Test Plan (on devserver):
```
$make db_stress
$./db_stress
$make all && make check
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5950
Differential Revision: D18044399
Pulled By: riversand963
fbshipit-source-id: 895585abbbdfd8b954965921dba4b1400b7af1b1
Summary:
Users may desire to specify extra dependencies via buck. This PR allows users to pass additional dependencies as a JSON object so that the buckifier script can generate TARGETS file with desired extra dependencies.
Test plan (on dev server)
```
$python buckifier/buckify_rocksdb.py '{"fake": {"extra_deps": [":test_dep", "//fakes/module:mock1"], "extra_compiler_flags": ["-DROCKSDB_LITE", "-Os"]}}'
Generating TARGETS
Extra dependencies:
{'': {'extra_compiler_flags': [], 'extra_deps': []}, 'test_dep1': {'extra_compiler_flags': ['-O2', '-DROCKSDB_LITE'], 'extra_deps': [':fake', '//dep1/mock']}}
Generated TARGETS Summary:
- 5 libs
- 0 binarys
- 296 tests
```
Verify the TARGETS file.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5648
Differential Revision: D16565043
Pulled By: riversand963
fbshipit-source-id: a6ef02274174fcf159692d7b846e828454d01e89
Summary:
Update buckifier templates in the scripts.
Test plan (on devserver)
```
$python buckifier/buckify_rocksdb.py
```
Then
```
$git diff
```
Verify that generated TARGETS file is the same (except for indentation).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5647
Differential Revision: D16555647
Pulled By: riversand963
fbshipit-source-id: 32574a4d0e820858eab2391304dd731141719bcd
Summary:
This change adds a Dynamic Library class to the RocksDB Env. Dynamic libraries are populated via the Env::LoadLibrary method.
The addition of dynamic library support allows for a few different features to be developed:
1. The compression code can be changed to use dynamic library support. This would allow RocksDB to determine at run-time what compression packages were installed. This change would eliminate the need to make sure the build-time and run-time environment had the same library set. It would also simplify some of the Java build issues (where it attempts to build and include various packages inside the RocksDB jars).
2. Along with other features (to be provided in a subsequent PR), this change would allow code/configurations to be added to RocksDB at run-time. For example, the build system includes code for building an "rados" environment and adding "Cassandra" features. Instead of these extensions being built into the base RocksDB code, these extensions could be loaded at run-time as required/appropriate, either by configuration or explicitly.
We intend to push out other changes in support of the extending RocksDB at run-time via configurations.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5281
Differential Revision: D15447613
Pulled By: riversand963
fbshipit-source-id: 452cd4f54511c0bceee18f6d9d919aae9fd25fef
Summary:
There are too many types of files under util/. Some test related files don't belong to there or just are just loosely related. Mo
ve them to a new directory test_util/, so that util/ is cleaner.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5377
Differential Revision: D15551366
Pulled By: siying
fbshipit-source-id: 0f5c8653832354ef8caa31749c0143815d719e2c
anand76