Summary:
Introducing WriteController, which is a source of truth about per-DB write delays. Let's define an DB epoch as a period where there are no flushes and compactions (i.e. new epoch is started when flush or compaction finishes). Each epoch can either:
* proceed with all writes without delay
* delay all writes by fixed time
* stop all writes
The three modes are recomputed at each epoch change (flush, compaction), rather than on every write (which is currently the case).
When we have a lot of column families, our current pull behavior adds a big overhead, since we need to loop over every column family for every write. With new push model, overhead on Write code-path is minimal.
This is just the start. Next step is to also take care of stalls introduced by slow memtable flushes. The final goal is to eliminate function MakeRoomForWrite(), which currently needs to be called for every column family by every write.
Test Plan: make check for now. I'll add some unit tests later. Also, perf test.
Reviewers: dhruba, yhchiang, MarkCallaghan, sdong, ljin
Reviewed By: ljin
Subscribers: leveldb
Differential Revision: https://reviews.facebook.net/D22791
Summary:
1. Make filter_block.h a base class. Derive block_based_filter_block and full_filter_block. The previous one is the traditional filter block. The full_filter_block is newly added. It would generate a filter block that contain all the keys in SST file.
2. When querying a key, table would first check if full_filter is available. If not, it would go to the exact data block and check using block_based filter.
3. User could choose to use full_filter or tradional(block_based_filter). They would be stored in SST file with different meta index name. "filter.filter_policy" or "full_filter.filter_policy". Then, Table reader is able to know the fllter block type.
4. Some optimizations have been done for full_filter_block, thus it requires a different interface compared to the original one in filter_policy.h.
5. Actual implementation of filter bits coding/decoding is placed in util/bloom_impl.cc
Benchmark: base commit 1d23b5c470
Command:
db_bench --db=/dev/shm/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --write_buffer_size=134217728 --max_write_buffer_number=2 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --verify_checksum=false --max_background_compactions=4 --use_plain_table=0 --memtablerep=prefix_hash --open_files=-1 --mmap_read=1 --mmap_write=0 --bloom_bits=10 --bloom_locality=1 --memtable_bloom_bits=500000 --compression_type=lz4 --num=393216000 --use_hash_search=1 --block_size=1024 --block_restart_interval=16 --use_existing_db=1 --threads=1 --benchmarks=readrandom —disable_auto_compactions=1
Read QPS increase for about 30% from 2230002 to 2991411.
Test Plan:
make all check
valgrind db_test
db_stress --use_block_based_filter = 0
./auto_sanity_test.sh
Reviewers: igor, yhchiang, ljin, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D20979
Summary: In DBImpl::Recover method, while loading memtables, also check if memtables are empty. Use this in DBImplReadonly to determine whether to lookup memtable or not.
Test Plan:
db_test
make check all
Reviewers: sdong, yhchiang, ljin, igor
Reviewed By: ljin
Subscribers: leveldb
Differential Revision: https://reviews.facebook.net/D22281
Summary:
Add WriteBatchWithIndex so that a user can query data out of a WriteBatch, to support MongoDB's read-its-own-write.
WriteBatchWithIndex uses a skiplist to store the binary index. The index stores the offset of the entry in the write batch. When searching for a key, the key for the entry is read by read the entry from the write batch from the offset.
Define a new iterator class for querying data out of WriteBatchWithIndex. A user can create an iterator of the write batch for one column family, seek to a key and keep calling Next() to see next entries.
I will add more unit tests if people are OK about this API.
Test Plan:
make all check
Add unit tests.
Reviewers: yhchiang, igor, MarkCallaghan, ljin
Reviewed By: ljin
Subscribers: dhruba, leveldb, xjin
Differential Revision: https://reviews.facebook.net/D21381
Summary: as title
Test Plan:
make all check
ran db_bench and saw seek stats at the end
Reviewers: yhchiang, igor, sdong
Reviewed By: sdong
Subscribers: leveldb
Differential Revision: https://reviews.facebook.net/D21651
Summary:
Contains the following changes:
- Implementation of cuckoo_table_factory
- Adding cuckoo table into AdaptiveTableFactory
- Adding cuckoo_table_db_test, similar to lines of plain_table_db_test
- Minor fixes to Reader: When a key is found in the table, return the key found instead of the search key.
- Minor fixes to Builder: Add table properties that are required by Version::UpdateTemporaryStats() during Get operation. Don't define curr_node as a reference variable as the memory locations may get reassigned during tree.push_back operation, leading to invalid memory access.
Test Plan:
cuckoo_table_reader_test --enable_perf
cuckoo_table_builder_test
cuckoo_table_db_test
make check all
make valgrind_check
make asan_check
Reviewers: sdong, igor, yhchiang, ljin
Reviewed By: ljin
Subscribers: leveldb
Differential Revision: https://reviews.facebook.net/D21219
* 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:
Contains:
- Implementation of TableReader based on Cuckoo Hashing
- Unittests for CuckooTableReader
- Performance test for TableReader
Test Plan:
make cuckoo_table_reader_test
./cuckoo_table_reader_test
make valgrind_check
make asan_check
Reviewers: yhchiang, sdong, igor, ljin
Reviewed By: ljin
Subscribers: leveldb
Differential Revision: https://reviews.facebook.net/D20511
Summary:
This diff is adding spatial index support to RocksDB.
When creating the DB user specifies a list of spatial indexes. Spatial indexes can cover different areas and have different resolution (i.e. number of tiles). This is useful for supporting different zoom levels.
Each element inserted into SpatialDB has:
* a bounding box, which determines how will the element be indexed
* string blob, which will usually be WKB representation of the polygon (http://en.wikipedia.org/wiki/Well-known_text)
* feature set, which is a map of key-value pairs, where value can be int, double, bool, null or a string. FeatureSet will be a set of tags associated with geo elements (for example, 'road': 'highway' and similar)
* a list of indexes to insert the element in. For example, small river element will be inserted in index for high zoom level, while country border will be inserted in all indexes (including the index for low zoom level).
Each query is executed on single spatial index. Query guarantees that it will return all elements intersecting the specified bounding box, but it might also return some extra non-intersecting elements.
Test Plan: Added bunch of unit tests in spatial_db_test
Reviewers: dhruba, yinwang
Reviewed By: yinwang
Subscribers: leveldb
Differential Revision: https://reviews.facebook.net/D20361
Summary:
Add the missing ROCKSDB_JAR variable in Makefile, which is mistakenly
removed in https://reviews.facebook.net/D20289.
Test Plan:
export ROCKSDB_JAR=
make rocksdbjava
Summary:
Add a function GetOptions(), where based on four parameters users give: read/write amplification threshold, memory budget for mem tables and target DB size, it picks up a compaction style and parameters for them. Background threads are not touched yet.
One limit of this algorithm: since compression rate and key/value size are hard to predict, it's hard to predict level 0 file size from write buffer size. Simply make 1:1 ratio here.
Sample results: https://reviews.facebook.net/P477
Test Plan: Will add some a unit test where some sample scenarios are given and see they pick the results that make sense
Reviewers: yhchiang, dhruba, haobo, igor, ljin
Reviewed By: ljin
Subscribers: leveldb
Differential Revision: https://reviews.facebook.net/D18741
Summary:
Fixed some make and linking issues of RocksDBJava. Specifically:
* Add JAVA_LDFLAGS, which does not include gflags
* rocksdbjava library now uses JAVA_LDFLAGS instead of LDFLAGS
* java/Makefile now includes build_config.mk
* rearrange make rocksdbjava workflow to ensure the library file is correctly
included in the jar file.
Test Plan:
make rocksdbjava
make jdb_bench
java/jdb_bench.sh
Reviewers: dhruba, swapnilghike, zzbennett, rsumbaly, ankgup87
Reviewed By: ankgup87
Subscribers: leveldb
Differential Revision: https://reviews.facebook.net/D20289
Summary:
Add -Wsign-compare to WARNING_FLAGS in Makefile as not all g++ compiler
include -Wsign-compare in -Wall when compiling '.h' file.
Test Plan: make -j32
Reviewers: ljin, igor, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D20169
Summary:
This is a rough sketch of our new document API. Would like to get some thoughts and comments about the high-level architecture and API.
I didn't optimize for performance at all. Leaving some low-hanging fruit so that we can be happy when we fix them! :)
Currently, bunch of features are not supported at all. Indexes can be only specified when creating database. There is no query planner whatsoever. This will all be added in due time.
Test Plan: Added a simple unit test
Reviewers: haobo, yhchiang, dhruba, sdong, ljin
Reviewed By: ljin
Subscribers: leveldb
Differential Revision: https://reviews.facebook.net/D18747
Summary:
A generic rate limiter that can be shared by threads and rocksdb
instances. Will use this to smooth out write traffic generated by
compaction and flush. This will help us get better p99 behavior on flash
storage.
Test Plan:
unit test output
==== Test RateLimiterTest.Rate
request size [1 - 1023], limit 10 KB/sec, actual rate: 10.374969 KB/sec, elapsed 2002265
request size [1 - 2047], limit 20 KB/sec, actual rate: 20.771242 KB/sec, elapsed 2002139
request size [1 - 4095], limit 40 KB/sec, actual rate: 41.285299 KB/sec, elapsed 2202424
request size [1 - 8191], limit 80 KB/sec, actual rate: 81.371605 KB/sec, elapsed 2402558
request size [1 - 16383], limit 160 KB/sec, actual rate: 162.541268 KB/sec, elapsed 3303500
Reviewers: yhchiang, igor, sdong
Reviewed By: sdong
Subscribers: leveldb
Differential Revision: https://reviews.facebook.net/D19359
Summary:
After evaluating options for JSON storage, I decided to implement our own. The reason is that we'll be able to optimize it better and we get to reduce unnecessary dependencies (which is what we'd get with folly).
I also plan to write a serializer/deserializer for JSONDocument with our own binary format similar to BSON. That way we'll store binary JSON format in RocksDB instead of the plain-text JSON. This means less storage and faster deserialization.
There are still some inefficiencies left here. I plan to optimize them after we develop a functioning DocumentDB. That way we can move and iterate faster.
Test Plan: added a unit test
Reviewers: dhruba, haobo, sdong, ljin, yhchiang
Reviewed By: haobo
Subscribers: leveldb
Differential Revision: https://reviews.facebook.net/D18831
Summary: We have a lot of problems with gflags. However, when compiling rocksdb static library, we don't need gflags dependency. Reorganize INSTALL.md such that first-time customers don't need any dependency installed to actually build rocksdb static library.
Test Plan: none
Reviewers: dhruba, haobo
Reviewed By: dhruba
CC: leveldb
Differential Revision: https://reviews.facebook.net/D18501
Summary:
db_test includes Benchmark for LogAndApply. This diff removes it from db_test and puts it into a separate log_and_apply bench. I just wanted to play around with our new benchmark framework and figure out how it works.
I would also like to show you how great it is! I believe right set of microbenchmarks can speed up our productivity a lot and help catch early regressions.
Test Plan: no
Reviewers: dhruba, haobo, sdong, ljin, yhchiang
Reviewed By: yhchiang
CC: leveldb
Differential Revision: https://reviews.facebook.net/D18261
Summary: Added benchmark functionality on the lines of folly/Benchmark.h
Test Plan: Added unit tests
Reviewers: igor, haobo, sdong, ljin, yhchiang, dhruba
Reviewed By: igor
CC: leveldb
Differential Revision: https://reviews.facebook.net/D17973
Summary:
The file tree structure in Version is prebuilt and the range of each file is known.
On the Get() code path, we do binary search in FindFile() by comparing
target key with each file's largest key and also check the range for each L0 file.
With some pre-calculated knowledge, each key comparision that has been done can serve
as a hint to narrow down further searches:
(1) If a key falls within a L0 file's range, we can safely skip the next
file if its range does not overlap with the current one.
(2) If a key falls within a file's range in level L0 - Ln-1, we should only
need to binary search in the next level for files that overlap with the current one.
(1) will be able to skip some files depending one the key distribution.
(2) can greatly reduce the range of binary search, especially for bottom
levels, given that one file most likely only overlaps with N files from
the level below (where N is max_bytes_for_level_multiplier). So on level
L, we will only look at ~N files instead of N^L files.
Some inital results: measured with 500M key DB, when write is light (10k/s = 1.2M/s), this
improves QPS ~7% on top of blocked bloom. When write is heavier (80k/s =
9.6M/s), it gives us ~13% improvement.
Test Plan: make all check
Reviewers: haobo, igor, dhruba, sdong, yhchiang
Reviewed By: haobo
CC: leveldb
Differential Revision: https://reviews.facebook.net/D17205
Summary:
Add a skeleton binding and test for BackupableDB which shows that BackupableDB
and RocksDB can share the same JNI calls.
Test Plan:
make rocksdbjava
make jtest
Reviewers: haobo, ankgup87, sdong, dhruba
Reviewed By: haobo
CC: leveldb
Differential Revision: https://reviews.facebook.net/D17793
Summary:
We don't really need sync_point.o if we're compiling with NDEBUG.
This diff depends on D17823
Test Plan: compiles
Reviewers: haobo, ljin, sdong
Reviewed By: haobo
CC: leveldb
Differential Revision: https://reviews.facebook.net/D17829
Summary:
Introducing RocksDBLite! Removes all the non-essential features and reduces the binary size. This effort should help our adoption on mobile.
Binary size when compiling for IOS (`TARGET_OS=IOS m static_lib`) is down to 9MB from 15MB (without stripping)
Test Plan: compiles :)
Reviewers: dhruba, haobo, ljin, sdong, yhchiang
Reviewed By: yhchiang
CC: leveldb
Differential Revision: https://reviews.facebook.net/D17835
Summary:
This is first step of my effort to reduce size of librocksdb.a for use in mobile.
ldb object files are huge and are ment to be used as a command line tool. I moved them to `tools/` directory and include them only when compiling `ldb`
This diff reduced librocksdb.a from 42MB to 39MB on my mac (not stripped).
Test Plan: ran ldb
Reviewers: dhruba, haobo, sdong, ljin, yhchiang
Reviewed By: yhchiang
CC: leveldb
Differential Revision: https://reviews.facebook.net/D17823
Summary:
* Add a benchmark for java binding for rocksdb. The java benchmark
is a complete rewrite based on the c++ db/db_bench.cc and the
DbBenchmark in dain's java leveldb.
* Support multithreading.
* 'readseq' is currently not supported as it requires RocksDB Iterator.
* usage:
--benchmarks
Comma-separated list of operations to run in the specified order
Actual benchmarks:
fillseq -- write N values in sequential key order in async mode
fillrandom -- write N values in random key order in async mode
fillbatch -- write N/1000 batch where each batch has 1000 values
in random key order in sync mode
fillsync -- write N/100 values in random key order in sync mode
fill100K -- write N/1000 100K values in random order in async mode
readseq -- read N times sequentially
readrandom -- read N times in random order
readhot -- read N times in random order from 1% section of DB
Meta Operations:
delete -- delete DB
DEFAULT: [fillseq, readrandom, fillrandom]
--compression_ratio
Arrange to generate values that shrink to this fraction of
their original size after compression
DEFAULT: 0.5
--use_existing_db
If true, do not destroy the existing database. If you set this
flag and also specify a benchmark that wants a fresh database, that benchmark will fail.
DEFAULT: false
--num
Number of key/values to place in database.
DEFAULT: 1000000
--threads
Number of concurrent threads to run.
DEFAULT: 1
--reads
Number of read operations to do. If negative, do --nums reads.
--key_size
The size of each key in bytes.
DEFAULT: 16
--value_size
The size of each value in bytes.
DEFAULT: 100
--write_buffer_size
Number of bytes to buffer in memtable before compacting
(initialized to default value by 'main'.)
DEFAULT: 4194304
--cache_size
Number of bytes to use as a cache of uncompressed data.
Negative means use default settings.
DEFAULT: -1
--seed
Seed base for random number generators.
DEFAULT: 0
--db
Use the db with the following name.
DEFAULT: /tmp/rocksdbjni-bench
* Add RocksDB.write().
Test Plan: make jbench
Reviewers: haobo, sdong, dhruba, ankgup87
Reviewed By: haobo
CC: leveldb
Differential Revision: https://reviews.facebook.net/D17433
