4 Commits (de1dd4ca190619ddef8517fb0667a894c00ae02b)

Author	SHA1	Message	Date
Alan Paxton	f8969ad7d4	Improve Java API get() performance by reducing copies (#10970) ... Summary: Performance improvements for `get()` paths in the RocksJava API (JNI). Document describing the performance results. Replace uses of the legacy `DB::Get()` method wrapper returning data in a `std::string` with direct calls to `DB::Get()` passing a pinnable slice to receive this data. Copying from a pinned slice direct to the destination java byte array, without going via an intervening std::string, is a major performance gain for this code path. Note that this gain only comes where `DB::Get()` is able to return a pinned buffer; where it has to copy into the buffer owned by the slice, there is still the intervening copy and no performance gain. It may be possible to address this case too, but it is not trivial. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10970 Reviewed By: pdillinger Differential Revision: D42125567 Pulled By: ajkr fbshipit-source-id: b7a4df7523b0420cadb1e9b6c7da3ec030a8da34	2 years ago
Alan Paxton	c1ec0b28eb	java / jni io_uring support (#9224) ... Summary: Existing multiGet() in java calls multi_get_helper() which then calls DB::std::vector MultiGet(). This doesn't take advantage of io_uring. This change adds another JNI level method that runs a parallel code path using the DB::void MultiGet(), using ByteBuffers at the JNI level. We call it multiGetDirect(). In addition to using the io_uring path, this code internally returns pinned slices which we can copy out of into our direct byte buffers; this should reduce the overall number of copies in the code path to/from Java. Some jmh benchmark runs (100k keys, 1000 key multiGet) suggest that for value sizes > 1k, we see about a 20% performance improvement, although performance is slightly reduced for small value sizes, there's a little bit more overhead in the JNI methods. Closes https://github.com/facebook/rocksdb/issues/8407 Pull Request resolved: https://github.com/facebook/rocksdb/pull/9224 Reviewed By: mrambacher Differential Revision: D32951754 Pulled By: jay-zhuang fbshipit-source-id: 1f70df7334be2b6c42a9c8f92725f67c71631690	3 years ago
Adam Retter	7242dae7fe	Improve RocksJava Comparator (#6252) ... Summary: This is a redesign of the API for RocksJava comparators with the aim of improving performance. It also simplifies the class hierarchy. **NOTE**: This breaks backwards compatibility for existing 3rd party Comparators implemented in Java... so we need to consider carefully which release branches this goes into. Previously when implementing a comparator in Java the developer had a choice of subclassing either `DirectComparator` or `Comparator` which would use direct and non-direct byte-buffers resepectively (via `DirectSlice` and `Slice`). In this redesign there we have eliminated the overhead of using the Java Slice classes, and just use `ByteBuffer`s. The `ComparatorOptions` supplied when constructing a Comparator allow you to choose between direct and non-direct byte buffers by setting `useDirect`. In addition, the `ComparatorOptions` now allow you to choose whether a ByteBuffer is reused over multiple comparator calls, by setting `maxReusedBufferSize > 0`. When buffers are reused, ComparatorOptions provides a choice of mutex type by setting `useAdaptiveMutex`. --- [JMH benchmarks previously indicated](https://github.com/facebook/rocksdb/pull/6241#issue-356398306) that the difference between C++ and Java for implementing a comparator was ~7x slowdown in Java. With these changes, when reusing buffers and guarding access to them via mutexes the slowdown is approximately the same. However, these changes offer a new facility to not reuse mutextes, which reduces the slowdown to ~5.5x in Java. We also offer a `thread_local` mechanism for reusing buffers, which reduces slowdown to ~5.2x in Java (closes https://github.com/facebook/rocksdb/pull/4425). These changes also form a good base for further optimisation work such as further JNI lookup caching, and JNI critical. --- These numbers were captured without jemalloc. With jemalloc, the performance improves for all tests, and the Java slowdown reduces to between 4.8x and 5.x. ``` ComparatorBenchmarks.put native_bytewise thrpt 25 124483.795 ± 2032.443 ops/s ComparatorBenchmarks.put native_reverse_bytewise thrpt 25 114414.536 ± 3486.156 ops/s ComparatorBenchmarks.put java_bytewise_non-direct_reused-64_adaptive-mutex thrpt 25 17228.250 ± 1288.546 ops/s ComparatorBenchmarks.put java_bytewise_non-direct_reused-64_non-adaptive-mutex thrpt 25 16035.865 ± 1248.099 ops/s ComparatorBenchmarks.put java_bytewise_non-direct_reused-64_thread-local thrpt 25 21571.500 ± 871.521 ops/s ComparatorBenchmarks.put java_bytewise_direct_reused-64_adaptive-mutex thrpt 25 23613.773 ± 8465.660 ops/s ComparatorBenchmarks.put java_bytewise_direct_reused-64_non-adaptive-mutex thrpt 25 16768.172 ± 5618.489 ops/s ComparatorBenchmarks.put java_bytewise_direct_reused-64_thread-local thrpt 25 23921.164 ± 8734.742 ops/s ComparatorBenchmarks.put java_bytewise_non-direct_no-reuse thrpt 25 17899.684 ± 839.679 ops/s ComparatorBenchmarks.put java_bytewise_direct_no-reuse thrpt 25 22148.316 ± 1215.527 ops/s ComparatorBenchmarks.put java_reverse_bytewise_non-direct_reused-64_adaptive-mutex thrpt 25 11311.126 ± 820.602 ops/s ComparatorBenchmarks.put java_reverse_bytewise_non-direct_reused-64_non-adaptive-mutex thrpt 25 11421.311 ± 807.210 ops/s ComparatorBenchmarks.put java_reverse_bytewise_non-direct_reused-64_thread-local thrpt 25 11554.005 ± 960.556 ops/s ComparatorBenchmarks.put java_reverse_bytewise_direct_reused-64_adaptive-mutex thrpt 25 22960.523 ± 1673.421 ops/s ComparatorBenchmarks.put java_reverse_bytewise_direct_reused-64_non-adaptive-mutex thrpt 25 18293.317 ± 1434.601 ops/s ComparatorBenchmarks.put java_reverse_bytewise_direct_reused-64_thread-local thrpt 25 24479.361 ± 2157.306 ops/s ComparatorBenchmarks.put java_reverse_bytewise_non-direct_no-reuse thrpt 25 7942.286 ± 626.170 ops/s ComparatorBenchmarks.put java_reverse_bytewise_direct_no-reuse thrpt 25 11781.955 ± 1019.843 ops/s ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/6252 Differential Revision: D19331064 Pulled By: pdillinger fbshipit-source-id: 1f3b794e6a14162b2c3ffb943e8c0e64a0c03738	5 years ago
Adam Retter	6477075f2c	JMH microbenchmarks for RocksJava (#6241) ... Summary: This is the start of some JMH microbenchmarks for RocksJava. Such benchmarks can help us decide on performance improvements of the Java API. At the moment, I have only added benchmarks for various Comparator options, as that is one of the first areas where I want to improve performance. I plan to expand this to many more tests. Details of how to compile and run the benchmarks are in the `README.md`. A run of these on a XEON 3.5 GHz 4vCPU (QEMU Virtual CPU version 2.5+) / 8GB RAM KVM with Ubuntu 18.04, OpenJDK 1.8.0_232, and gcc 8.3.0 produced the following: ``` # Run complete. Total time: 01:43:17 REMEMBER: The numbers below are just data. To gain reusable insights, you need to follow up on why the numbers are the way they are. Use profilers (see -prof, -lprof), design factorial experiments, perform baseline and negative tests that provide experimental control, make sure the benchmarking environment is safe on JVM/OS/HW level, ask for reviews from the domain experts. Do not assume the numbers tell you what you want them to tell. Benchmark (comparatorName) Mode Cnt Score Error Units ComparatorBenchmarks.put native_bytewise thrpt 25 122373.920 ± 2200.538 ops/s ComparatorBenchmarks.put java_bytewise_adaptive_mutex thrpt 25 17388.201 ± 1444.006 ops/s ComparatorBenchmarks.put java_bytewise_non-adaptive_mutex thrpt 25 16887.150 ± 1632.204 ops/s ComparatorBenchmarks.put java_direct_bytewise_adaptive_mutex thrpt 25 15644.572 ± 1791.189 ops/s ComparatorBenchmarks.put java_direct_bytewise_non-adaptive_mutex thrpt 25 14869.601 ± 2252.135 ops/s ComparatorBenchmarks.put native_reverse_bytewise thrpt 25 116528.735 ± 4168.797 ops/s ComparatorBenchmarks.put java_reverse_bytewise_adaptive_mutex thrpt 25 10651.975 ± 545.998 ops/s ComparatorBenchmarks.put java_reverse_bytewise_non-adaptive_mutex thrpt 25 10514.224 ± 930.069 ops/s ``` Indicating a ~7x difference between comparators implemented natively (C++) and those implemented in Java. Let's see if we can't improve on that in the near future... Pull Request resolved: https://github.com/facebook/rocksdb/pull/6241 Differential Revision: D19290410 Pulled By: pdillinger fbshipit-source-id: 25d44bf3a31de265502ed0c5d8a28cf4c7cb9c0b	5 years ago