基准测试工具

db_bench

db_bench是用来测试RocksDB性能的主要工具。RocksDB继承了LevelDB中的db_bench，并对其进行了增强，以支持许多其他选项。 db_bench支持许多基准测试来生成不同类型的工作负载，它的各种选项可用于控制测试。

如果您刚刚开始使用db_bench，下面是一些您可以尝试的事情:

1.从一个简单的基准测试(如fillseq(或fillrandom))开始，创建一个数据库并用一些数据填充它

  1. ./db_bench --benchmarks="fillseq"

如果您想要更多的统计信息，请添加元操作符”stats”和—statistics标志。

  1. ./db_bench --benchmarks="fillseq,stats" --statistics

2.读取数据

  1. ./db_bench --benchmarks="readrandom" --use_existing_db

您还可以将多个基准测试组合到传递给的字符串— 基准测试以便它们按顺序运行。例子:

  1. ./db_bench --benchmarks="fillseq,readrandom,readseq"

db_bench用法的更深入示例可以在 这里(https://github.com/facebook/rocksdb/wiki/performance-benchmarks) 和 这里(https://github.com/facebook/rocksdb/wiki/RocksDB-In-Memory-Workload-Performance-Benchmarks) 找到。

标准列表:

  1. fillseq       -- write N values in sequential key order in async mode
  2. fillseqdeterministic       -- write N values in the specified key order and keep the shape of the LSM tree
  3. fillrandom    -- write N values in random key order in async mode
  4. filluniquerandomdeterministic       -- write N values in a random key order and keep the shape of the LSM tree
  5. overwrite     -- overwrite N values in random key order in async mode
  6. fillsync      -- write N/100 values in random key order in sync mode
  7. fill100K      -- write N/1000 100K values in random order in async mode
  8. deleteseq     -- delete N keys in sequential order
  9. deleterandom  -- delete N keys in random order
  10. readseq       -- read N times sequentially
  11. readtocache   -- 1 thread reading database sequentially
  12. readreverse   -- read N times in reverse order
  13. readrandom    -- read N times in random order
  14. readmissing   -- read N missing keys in random order
  15. readwhilewriting      -- 1 writer, N threads doing random reads
  16. readwhilemerging      -- 1 merger, N threads doing random reads
  17. readrandomwriterandom -- N threads doing random-read, random-write
  18. prefixscanrandom      -- prefix scan N times in random order
  19. updaterandom  -- N threads doing read-modify-write for random keys
  20. appendrandom  -- N threads doing read-modify-write with growing values
  21. mergerandom   -- same as updaterandom/appendrandom using merge operator. Must be used with merge_operator
  22. readrandommergerandom -- perform N random read-or-merge operations. Must be used with merge_operator
  23. newiterator   -- repeated iterator creation
  24. seekrandom    -- N random seeks, call Next seek_nexts times per seek
  25. seekrandomwhilewriting -- seekrandom and 1 thread doing overwrite
  26. seekrandomwhilemerging -- seekrandom and 1 thread doing merge
  27. crc32c        -- repeated crc32c of 4K of data
  28. xxhash        -- repeated xxHash of 4K of data
  29. acquireload   -- load N*1000 times
  30. fillseekseq   -- write N values in sequential key, then read them by seeking to each key
  31. randomtransaction     -- execute N random transactions and verify correctness
  32. randomreplacekeys     -- randomly replaces N keys by deleting the old version and putting the new version
  33. timeseries            -- 1 writer generates time series data and multiple readers doing random reads on id

所有选项的列表:

  1. $ ./db_bench -help
  2. db_bench:
  3. USAGE:
  4. ./db_bench [OPTIONS]...
  6.   Flags from tools/db_bench_tool.cc:
  7.     -advise_random_on_open (Advise random access on table file open) type: bool
  8.       default: true
  9.     -allow_concurrent_memtable_write (Allow multi-writers to update mem tables
  10.       in parallel.) type: bool default: false
  11.     -base_background_compactions (The base number of concurrent background
  12.       compactions to occur in parallel.) type: int32 default: 1
  13.     -batch_size (Batch size) type: int64 default: 1
  14.     -benchmark_read_rate_limit (If non-zero, db_bench will rate-limit the reads
  15.       from RocksDB. This is the global rate in ops/second.) type: uint64
  16.       default: 0
  17.     -benchmark_write_rate_limit (If non-zero, db_bench will rate-limit the
  18.       writes going into RocksDB. This is the global rate in bytes/second.)
  19.       type: uint64 default: 0
  20.     -benchmarks (Comma-separated list of operations to run in the specified
  21.       order. Available benchmarks:
  22.         fillseq       -- write N values in sequential key order in async mode
  23.         fillseqdeterministic       -- write N values in the specified key order
  24.       and keep the shape of the LSM tree
  25.         fillrandom    -- write N values in random key order in async mode
  26.         filluniquerandomdeterministic       -- write N values in a random key
  27.       order and keep the shape of the LSM tree
  28.         overwrite     -- overwrite N values in random key order in async mode
  29.         fillsync      -- write N/100 values in random key order in sync mode
  30.         fill100K      -- write N/1000 100K values in random order in async mode
  31.         deleteseq     -- delete N keys in sequential order
  32.         deleterandom  -- delete N keys in random order
  33.         readseq       -- read N times sequentially
  34.         readtocache   -- 1 thread reading database sequentially
  35.         readreverse   -- read N times in reverse order
  36.         readrandom    -- read N times in random order
  37.         readmissing   -- read N missing keys in random order
  38.         readwhilewriting      -- 1 writer, N threads doing random reads
  39.         readwhilemerging      -- 1 merger, N threads doing random reads
  40.         readrandomwriterandom -- N threads doing random-read, random-write
  41.         prefixscanrandom      -- prefix scan N times in random order
  42.         updaterandom  -- N threads doing read-modify-write for random keys
  43.         appendrandom  -- N threads doing read-modify-write with growing values
  44.         mergerandom   -- same as updaterandom/appendrandom using merge operator.
  45.       Must be used with merge_operator
  46.         readrandommergerandom -- perform N random read-or-merge operations. Must
  47.       be used with merge_operator
  48.         newiterator   -- repeated iterator creation
  49.         seekrandom    -- N random seeks, call Next seek_nexts times per seek
  50.         seekrandomwhilewriting -- seekrandom and 1 thread doing overwrite
  51.         seekrandomwhilemerging -- seekrandom and 1 thread doing merge
  52.         crc32c        -- repeated crc32c of 4K of data
  53.         xxhash        -- repeated xxHash of 4K of data
  54.         acquireload   -- load N*1000 times
  55.         fillseekseq   -- write N values in sequential key, then read them by
  56.       seeking to each key
  57.         randomtransaction     -- execute N random transactions and verify
  58.       correctness
  59.         randomreplacekeys     -- randomly replaces N keys by deleting the old
  60.       version and putting the new version
  62.         timeseries            -- 1 writer generates time series data and
  63.       multiple readers doing random reads on id
  65.       Meta operations:
  66.         compact     -- Compact the entire DB
  67.         stats       -- Print DB stats
  68.         resetstats  -- Reset DB stats
  69.         levelstats  -- Print the number of files and bytes per level
  70.         sstables    -- Print sstable info
  71.         heapprofile -- Dump a heap profile (if supported by this port)
  72.       ) type: string
  73.       default: "fillseq,fillseqdeterministic,fillsync,fillrandom,filluniquerandomdeterministic,overwrite,readrandom,newiterator,newiteratorwhilewriting,seekrandom,seekrandomwhilewriting,seekrandomwhilemerging,readseq,readreverse,compact,readrandom,multireadrandom,readseq,readtocache,readreverse,readwhilewriting,readwhilemerging,readrandomwriterandom,updaterandom,randomwithverify,fill100K,crc32c,xxhash,compress,uncompress,acquireload,fillseekseq,randomtransaction,randomreplacekeys,timeseries"
  74.     -block_restart_interval (Number of keys between restart points for delta
  75.       encoding of keys in data block.) type: int32 default: 16
  76.     -block_size (Number of bytes in a block.) type: int32 default: 4096
  77.     -bloom_bits (Bloom filter bits per key. Negative means use default
  78.       settings.) type: int32 default: -1
  79.     -bloom_locality (Control bloom filter probes locality) type: int32
  80.       default: 0
  81.     -bytes_per_sync (Allows OS to incrementally sync SST files to disk while
  82.       they are being written, in the background. Issue one request for every
  83.       bytes_per_sync written. 0 turns it off.) type: uint64 default: 0
  84.     -cache_high_pri_pool_ratio (Ratio of block cache reserve for high pri
  85.       blocks. If > 0.0, we also enable
  86.       cache_index_and_filter_blocks_with_high_priority.) type: double
  87.       default: 0
  88.     -cache_index_and_filter_blocks (Cache index/filter blocks in block cache.)
  89.       type: bool default: false
  90.     -cache_numshardbits (Number of shards for the block cache is 2 **
  91.       cache_numshardbits. Negative means use default settings. This is applied
  92.       only if FLAGS_cache_size is non-negative.) type: int32 default: 6
  93.     -cache_size (Number of bytes to use as a cache of uncompressed data)
  94.       type: int64 default: 8388608
  95.     -compaction_fadvice (Access pattern advice when a file is compacted)
  96.       type: string default: "NORMAL"
  97.     -compaction_pri (priority of files to compaction: by size or by data age)
  98.       type: int32 default: 0
  99.     -compaction_readahead_size (Compaction readahead size) type: int32
  100.       default: 0
  101.     -compaction_style (style of compaction: level-based, universal and fifo)
  102.       type: int32 default: 0
  103.     -compressed_cache_size (Number of bytes to use as a cache of compressed
  104.       data.) type: int64 default: -1
  105.     -compression_level (Compression level. For zlib this should be -1 for the
  106.       default level, or between 0 and 9.) type: int32 default: -1
  107.     -compression_max_dict_bytes (Maximum size of dictionary used to prime the
  108.       compression library.) type: int32 default: 0
  109.     -compression_ratio (Arrange to generate values that shrink to this fraction
  110.       of their original size after compression) type: double default: 0.5
  111.     -compression_type (Algorithm to use to compress the database) type: string
  112.       default: "snappy"
  113.     -cuckoo_hash_ratio (Hash ratio for Cuckoo SST table.) type: double
  114.       default: 0.90000000000000002
  115.     -db (Use the db with the following name.) type: string default: ""
  116.     -db_write_buffer_size (Number of bytes to buffer in all memtables before
  117.       compacting) type: int64 default: 0
  118.     -delayed_write_rate (Limited bytes allowed to DB when soft_rate_limit or
  119.       level0_slowdown_writes_trigger triggers) type: uint64 default: 8388608
  120.     -delete_obsolete_files_period_micros (Ignored. Left here for backward
  121.       compatibility) type: uint64 default: 0
  122.     -deletepercent (Percentage of deletes out of reads/writes/deletes (used in
  123.       RandomWithVerify only). RandomWithVerify calculates writepercent as (100
  124.       - FLAGS_readwritepercent - deletepercent), so deletepercent must be
  125.       smaller than (100 - FLAGS_readwritepercent)) type: int32 default: 2
  126.     -deletes (Number of delete operations to do.  If negative, do FLAGS_num
  127.       deletions.) type: int64 default: -1
  128.     -disable_auto_compactions (Do not auto trigger compactions) type: bool
  129.       default: false
  130.     -disable_seek_compaction (Not used, left here for backwards compatibility)
  131.       type: int32 default: 0
  132.     -disable_wal (If true, do not write WAL for write.) type: bool
  133.       default: false
  134.     -dump_malloc_stats (Dump malloc stats in LOG ) type: bool default: true
  135.     -duration (Time in seconds for the random-ops tests to run. When 0 then num
  136.       & reads determine the test duration) type: int32 default: 0
  137.     -enable_io_prio (Lower the background flush/compaction threads' IO
  138.       priority) type: bool default: false
  139.     -enable_numa (Make operations aware of NUMA architecture and bind memory
  140.       and cpus corresponding to nodes together. In NUMA, memory in same node as
  141.       CPUs are closer when compared to memory in other nodes. Reads can be
  142.       faster when the process is bound to CPU and memory of same node. Use
  143.       "$numactl --hardware" command to see NUMA memory architecture.)
  144.       type: bool default: false
  145.     -enable_pipelined_write (Allow WAL and memtable writes to be pipelined)
  146.       type: bool default: true
  147.     -enable_write_thread_adaptive_yield (Use a yielding spin loop for brief
  148.       writer thread waits.) type: bool default: false
  149.     -env_uri (URI for registry Env lookup. Mutually exclusive with --hdfs.)
  150.       type: string default: ""
  151.     -expand_range_tombstones (Expand range tombstone into sequential regular
  152.       tombstones.) type: bool default: false
  153.     -expire_style (Style to remove expired time entries. Can be one of the
  154.       options below: none (do not expired data), compaction_filter (use a
  155.       compaction filter to remove expired data), delete (seek IDs and remove
  156.       expired data) (used in TimeSeries only).) type: string default: "none"
  157.     -fifo_compaction_allow_compaction (Allow compaction in FIFO compaction.)
  158.       type: bool default: true
  159.     -fifo_compaction_max_table_files_size_mb (The limit of total table file
  160.       sizes to trigger FIFO compaction) type: uint64 default: 0
  161.     -file_opening_threads (If open_files is set to -1, this option set the
  162.       number of threads that will be used to open files during DB::Open())
  163.       type: int32 default: 16
  164.     -finish_after_writes (Write thread terminates after all writes are
  165.       finished) type: bool default: false
  166.     -hard_pending_compaction_bytes_limit (Stop writes if pending compaction
  167.       bytes exceed this number) type: uint64 default: 137438953472
  168.     -hard_rate_limit (DEPRECATED) type: double default: 0
  169.     -hash_bucket_count (hash bucket count) type: int64 default: 1048576
  170.     -hdfs (Name of hdfs environment. Mutually exclusive with --env_uri.)
  171.       type: string default: ""
  172.     -histogram (Print histogram of operation timings) type: bool default: false
  173.     -identity_as_first_hash (the first hash function of cuckoo table becomes an
  174.       identity function. This is only valid when key is 8 bytes) type: bool
  175.       default: false
  176.     -index_block_restart_interval (Number of keys between restart points for
  177.       delta encoding of keys in index block.) type: int32 default: 1
  178.     -key_id_range (Range of possible value of key id (used in TimeSeries
  179.       only).) type: int32 default: 100000
  180.     -key_size (size of each key) type: int32 default: 16
  181.     -keys_per_prefix (control average number of keys generated per prefix, 0
  182.       means no special handling of the prefix, i.e. use the prefix comes with
  183.       the generated random number.) type: int64 default: 0
  184.     -level0_file_num_compaction_trigger (Number of files in level-0 when
  185.       compactions start) type: int32 default: 4
  186.     -level0_slowdown_writes_trigger (Number of files in level-0 that will slow
  187.       down writes.) type: int32 default: 20
  188.     -level0_stop_writes_trigger (Number of files in level-0 that will trigger
  189.       put stop.) type: int32 default: 36
  190.     -level_compaction_dynamic_level_bytes (Whether level size base is dynamic)
  191.       type: bool default: false
  192.     -max_background_compactions (The maximum number of concurrent background
  193.       compactions that can occur in parallel.) type: int32 default: 1
  194.     -max_background_flushes (The maximum number of concurrent background
  195.       flushes that can occur in parallel.) type: int32 default: 1
  196.     -max_bytes_for_level_base (Max bytes for level-1) type: uint64
  197.       default: 268435456
  198.     -max_bytes_for_level_multiplier (A multiplier to compute max bytes for
  199.       level-N (N >= 2)) type: double default: 10
  200.     -max_bytes_for_level_multiplier_additional (A vector that specifies
  201.       additional fanout per level) type: string default: ""
  202.     -max_compaction_bytes (Max bytes allowed in one compaction) type: uint64
  203.       default: 0
  204.     -max_num_range_tombstones (Maximum number of range tombstones to insert.)
  205.       type: int64 default: 0
  206.     -max_successive_merges (Maximum number of successive merge operations on a
  207.       key in the memtable) type: int32 default: 0
  208.     -max_total_wal_size (Set total max WAL size) type: uint64 default: 0
  209.     -max_write_buffer_number (The number of in-memory memtables. Each memtable
  210.       is of sizewrite_buffer_size.) type: int32 default: 2
  211.     -max_write_buffer_number_to_maintain (The total maximum number of write
  212.       buffers to maintain in memory including copies of buffers that have
  213.       already been flushed. Unlike max_write_buffer_number, this parameter does
  214.       not affect flushing. This controls the minimum amount of write history
  215.       that will be available in memory for conflict checking when Transactions
  216.       are used. If this value is too low, some transactions may fail at commit
  217.       time due to not being able to determine whether there were any write
  218.       conflicts. Setting this value to 0 will cause write buffers to be freed
  219.       immediately after they are flushed.  If this value is set to -1,
  220.       'max_write_buffer_number' will be used.) type: int32 default: 0
  221.     -memtable_bloom_size_ratio (Ratio of memtable size used for bloom filter. 0
  222.       means no bloom filter.) type: double default: 0
  223.     -memtable_insert_with_hint_prefix_size (If non-zero, enable memtable insert
  224.       with hint with the given prefix size.) type: int32 default: 0
  225.     -memtable_use_huge_page (Try to use huge page in memtables.) type: bool
  226.       default: false
  227.     -memtablerep () type: string default: "skip_list"
  228.     -merge_keys (Number of distinct keys to use for MergeRandom and
  229.       ReadRandomMergeRandom. If negative, there will be FLAGS_num keys.)
  230.       type: int64 default: -1
  231.     -merge_operator (The merge operator to use with the database.If a new merge
  232.       operator is specified, be sure to use fresh database The possible merge
  233.       operators are defined in utilities/merge_operators.h) type: string
  234.       default: ""
  235.     -mergereadpercent (Ratio of merges to merges&reads (expressed as
  236.       percentage) for the ReadRandomMergeRandom workload. The default value 70
  237.       means 70% out of all read and merge operations are merges. In other
  238.       words, 7 merges for every 3 gets.) type: int32 default: 70
  239.     -min_level_to_compress (If non-negative, compression starts from this
  240.       level. Levels with number < min_level_to_compress are not compressed.
  241.       Otherwise, apply compression_type to all levels.) type: int32 default: -1
  242.     -min_write_buffer_number_to_merge (The minimum number of write buffers that
  243.       will be merged togetherbefore writing to storage. This is cheap because
  244.       it is anin-memory merge. If this feature is not enabled, then all
  245.       thesewrite buffers are flushed to L0 as separate files and this increases
  246.       read amplification because a get request has to check in all of these
  247.       files. Also, an in-memory merge may result in writing less data to
  248.       storage if there are duplicate records  in each of these individual write
  249.       buffers.) type: int32 default: 1
  250.     -mmap_read (Allow reads to occur via mmap-ing files) type: bool
  251.       default: false
  252.     -mmap_write (Allow writes to occur via mmap-ing files) type: bool
  253.       default: false
  254.     -new_table_reader_for_compaction_inputs (If true, uses a separate file
  255.       handle for compaction inputs) type: int32 default: 1
  256.     -num (Number of key/values to place in database) type: int64
  257.       default: 1000000
  258.     -num_column_families (Number of Column Families to use.) type: int32
  259.       default: 1
  260.     -num_deletion_threads (Number of threads to do deletion (used in TimeSeries
  261.       and delete expire_style only).) type: int32 default: 1
  262.     -num_hot_column_families (Number of Hot Column Families. If more than 0,
  263.       only write to this number of column families. After finishing all the
  264.       writes to them, create new set of column families and insert to them.
  265.       Only used when num_column_families > 1.) type: int32 default: 0
  266.     -num_levels (The total number of levels) type: int32 default: 7
  267.     -num_multi_db (Number of DBs used in the benchmark. 0 means single DB.)
  268.       type: int32 default: 0
  269.     -numdistinct (Number of distinct keys to use. Used in RandomWithVerify to
  270.       read/write on fewer keys so that gets are more likely to find the key and
  271.       puts are more likely to update the same key) type: int64 default: 1000
  272.     -open_files (Maximum number of files to keep open at the same time (use
  273.       default if == 0)) type: int32 default: -1
  274.     -optimistic_transaction_db (Open a OptimisticTransactionDB instance.
  275.       Required for randomtransaction benchmark.) type: bool default: false
  276.     -optimize_filters_for_hits (Optimizes bloom filters for workloads for most
  277.       lookups return a value. For now this doesn't create bloom filters for the
  278.       max level of the LSM to reduce metadata that should fit in RAM. )
  279.       type: bool default: false
  280.     -options_file (The path to a RocksDB options file.  If specified, then
  281.       db_bench will run with the RocksDB options in the default column family
  282.       of the specified options file. Note that with this setting, db_bench will
  283.       ONLY accept the following RocksDB options related command-line arguments,
  284.       all other arguments that are related to RocksDB options will be ignored:
  285.         --use_existing_db
  286.         --statistics
  287.         --row_cache_size
  288.         --row_cache_numshardbits
  289.         --enable_io_prio
  290.         --dump_malloc_stats
  291.         --num_multi_db
  292.       ) type: string default: ""
  293.     -perf_level (Level of perf collection) type: int32 default: 1
  294.     -pin_l0_filter_and_index_blocks_in_cache (Pin index/filter blocks of L0
  295.       files in block cache.) type: bool default: false
  296.     -pin_slice (use pinnable slice for point lookup) type: bool default: true
  297.     -prefix_size (control the prefix size for HashSkipList and plain table)
  298.       type: int32 default: 0
  299.     -random_access_max_buffer_size (Maximum windows randomaccess buffer size)
  300.       type: int32 default: 1048576
  301.     -range_tombstone_width (Number of keys in tombstone's range) type: int64
  302.       default: 100
  303.     -rate_limit_delay_max_milliseconds (When hard_rate_limit is set then this
  304.       is the max time a put will be stalled.) type: int32 default: 1000
  305.     -rate_limiter_bytes_per_sec (Set options.rate_limiter value.) type: uint64
  306.       default: 0
  307.     -read_amp_bytes_per_bit (Number of bytes per bit to be used in block
  308.       read-amp bitmap) type: int32 default: 0
  309.     -read_cache_direct_read (Whether to use Direct IO for reading from read
  310.       cache) type: bool default: true
  311.     -read_cache_direct_write (Whether to use Direct IO for writing to the read
  312.       cache) type: bool default: true
  313.     -read_cache_path (If not empty string, a read cache will be used in this
  314.       path) type: string default: ""
  315.     -read_cache_size (Maximum size of the read cache) type: int64
  316.       default: 4294967296
  317.     -read_random_exp_range (Read random's key will be generated using
  318.       distribution of num * exp(-r) where r is uniform number from 0 to this
  319.       value. The larger the number is, the more skewed the reads are. Only used
  320.       in readrandom and multireadrandom benchmarks.) type: double default: 0
  321.     -readonly (Run read only benchmarks.) type: bool default: false
  322.     -reads (Number of read operations to do.  If negative, do FLAGS_num reads.)
  323.       type: int64 default: -1
  324.     -readwritepercent (Ratio of reads to reads/writes (expressed as percentage)
  325.       for the ReadRandomWriteRandom workload. The default value 90 means 90%
  326.       operations out of all reads and writes operations are reads. In other
  327.       words, 9 gets for every 1 put.) type: int32 default: 90
  328.     -report_bg_io_stats (Measure times spents on I/Os while in compactions. )
  329.       type: bool default: false
  330.     -report_file (Filename where some simple stats are reported to (if
  331.       --report_interval_seconds is bigger than 0)) type: string
  332.       default: "report.csv"
  333.     -report_file_operations (if report number of file operations) type: bool
  334.       default: false
  335.     -report_interval_seconds (If greater than zero, it will write simple stats
  336.       in CVS format to --report_file every N seconds) type: int64 default: 0
  337.     -reverse_iterator (When true use Prev rather than Next for iterators that
  338.       do Seek and then Next) type: bool default: false
  339.     -row_cache_size (Number of bytes to use as a cache of individual rows (0 =
  340.       disabled).) type: int64 default: 0
  341.     -seed (Seed base for random number generators. When 0 it is deterministic.)
  342.       type: int64 default: 0
  343.     -seek_nexts (How many times to call Next() after Seek() in fillseekseq,
  344.       seekrandom, seekrandomwhilewriting and seekrandomwhilemerging)
  345.       type: int32 default: 0
  346.     -show_table_properties (If true, then per-level table properties will be
  347.       printed on every stats-interval when stats_interval is set and
  348.       stats_per_interval is on.) type: bool default: false
  349.     -simcache_size (Number of bytes to use as a simcache of uncompressed data.
  350.       Nagative value disables simcache.) type: int64 default: -1
  351.     -skip_list_lookahead (Used with skip_list memtablerep; try linear search
  352.       first for this many steps from the previous position) type: int32
  353.       default: 0
  354.     -soft_pending_compaction_bytes_limit (Slowdown writes if pending compaction
  355.       bytes exceed this number) type: uint64 default: 68719476736
  356.     -soft_rate_limit (DEPRECATED) type: double default: 0
  357.     -statistics (Database statistics) type: bool default: false
  358.     -statistics_string (Serialized statistics string) type: string default: ""
  359.     -stats_interval (Stats are reported every N operations when this is greater
  360.       than zero. When 0 the interval grows over time.) type: int64 default: 0
  361.     -stats_interval_seconds (Report stats every N seconds. This overrides
  362.       stats_interval when both are > 0.) type: int64 default: 0
  363.     -stats_per_interval (Reports additional stats per interval when this is
  364.       greater than 0.) type: int32 default: 0
  365.     -stddev (Standard deviation of normal distribution used for picking keys
  366.       (used in RandomReplaceKeys only).) type: double default: 2000
  367.     -subcompactions (Maximum number of subcompactions to divide L0-L1
  368.       compactions into.) type: uint64 default: 1
  369.     -sync (Sync all writes to disk) type: bool default: false
  370.     -table_cache_numshardbits () type: int32 default: 4
  371.     -target_file_size_base (Target file size at level-1) type: int64
  372.       default: 67108864
  373.     -target_file_size_multiplier (A multiplier to compute target level-N file
  374.       size (N >= 2)) type: int32 default: 1
  375.     -thread_status_per_interval (Takes and report a snapshot of the current
  376.       status of each thread when this is greater than 0.) type: int32
  377.       default: 0
  378.     -threads (Number of concurrent threads to run.) type: int32 default: 1
  379.     -time_range (Range of timestamp that store in the database (used in
  380.       TimeSeries only).) type: uint64 default: 100000
  381.     -transaction_db (Open a TransactionDB instance. Required for
  382.       randomtransaction benchmark.) type: bool default: false
  383.     -transaction_lock_timeout (If using a transaction_db, specifies the lock
  384.       wait timeout in milliseconds before failing a transaction waiting on a
  385.       lock) type: uint64 default: 100
  386.     -transaction_set_snapshot (Setting to true will have each transaction call
  387.       SetSnapshot() upon creation.) type: bool default: false
  388.     -transaction_sets (Number of keys each transaction will modify (use in
  389.       RandomTransaction only).  Max: 9999) type: uint64 default: 2
  390.     -transaction_sleep (Max microseconds to sleep in between reading and
  391.       writing a value (used in RandomTransaction only). ) type: int32
  392.       default: 0
  393.     -truth_db (Truth key/values used when using verify) type: string
  394.       default: "/dev/shm/truth_db/dbbench"
  395.     -universal_allow_trivial_move (Allow trivial move in universal compaction.)
  396.       type: bool default: false
  397.     -universal_compression_size_percent (The percentage of the database to
  398.       compress for universal compaction. -1 means compress everything.)
  399.       type: int32 default: -1
  400.     -universal_max_merge_width (The max number of files to compact in universal
  401.       style compaction) type: int32 default: 0
  402.     -universal_max_size_amplification_percent (The max size amplification for
  403.       universal style compaction) type: int32 default: 0
  404.     -universal_min_merge_width (The minimum number of files in a single
  405.       compaction run (for universal compaction only).) type: int32 default: 0
  406.     -universal_size_ratio (Percentage flexibility while comparing file size
  407.       (for universal compaction only).) type: int32 default: 0
  408.     -use_adaptive_mutex (Use adaptive mutex) type: bool default: false
  409.     -use_blob_db (Open a BlobDB instance. Required for largevalue benchmark.)
  410.       type: bool default: false
  411.     -use_block_based_filter (if use kBlockBasedFilter instead of kFullFilter
  412.       for filter block. This is valid if only we use BlockTable) type: bool
  413.       default: false
  414.     -use_clock_cache (Replace default LRU block cache with clock cache.)
  415.       type: bool default: false
  416.     -use_cuckoo_table (if use cuckoo table format) type: bool default: false
  417.     -use_direct_io_for_flush_and_compaction (Use O_DIRECT for background flush
  418.       and compaction I/O) type: bool default: false
  419.     -use_direct_reads (Use O_DIRECT for reading data) type: bool default: false
  420.     -use_existing_db (If true, do not destroy the existing database.  If you
  421.       set this flag and also specify a benchmark that wants a fresh database,
  422.       that benchmark will fail.) type: bool default: false
  423.     -use_fsync (If true, issue fsync instead of fdatasync) type: bool
  424.       default: false
  425.     -use_hash_search (if use kHashSearch instead of kBinarySearch. This is
  426.       valid if only we use BlockTable) type: bool default: false
  427.     -use_plain_table (if use plain table instead of block-based table format)
  428.       type: bool default: false
  429.     -use_single_deletes (Use single deletes (used in RandomReplaceKeys only).)
  430.       type: bool default: true
  431.     -use_stderr_info_logger (Write info logs to stderr instead of to LOG file.
  432.       ) type: bool default: false
  433.     -use_tailing_iterator (Use tailing iterator to access a series of keys
  434.       instead of get) type: bool default: false
  435.     -use_uint64_comparator (use Uint64 user comparator) type: bool
  436.       default: false
  437.     -value_size (Size of each value) type: int32 default: 100
  438.     -verify_checksum (Verify checksum for every block read from storage)
  439.       type: bool default: false
  440.     -wal_bytes_per_sync (Allows OS to incrementally sync WAL files to disk
  441.       while they are being written, in the background. Issue one request for
  442.       every wal_bytes_per_sync written. 0 turns it off.) type: uint64
  443.       default: 0
  444.     -wal_dir (If not empty, use the given dir for WAL) type: string default: ""
  445.     -wal_size_limit_MB (Set the size limit for the WAL Files in MB.)
  446.       type: uint64 default: 0
  447.     -wal_ttl_seconds (Set the TTL for the WAL Files in seconds.) type: uint64
  448.       default: 0
  449.     -writable_file_max_buffer_size (Maximum write buffer for Writable File)
  450.       type: int32 default: 1048576
  451.     -write_buffer_size (Number of bytes to buffer in memtable before
  452.       compacting) type: int64 default: 67108864
  453.     -write_thread_max_yield_usec (Maximum microseconds for
  454.       enable_write_thread_adaptive_yield operation.) type: uint64 default: 100
  455.     -write_thread_slow_yield_usec (The threshold at which a slow yield is
  456.       considered a signal that other processes or threads want the core.)
  457.       type: uint64 default: 3
  458.     -writes (Number of write operations to do. If negative, do --num reads.)
  459.       type: int64 default: -1
  460.     -writes_per_range_tombstone (Number of writes between range tombstones)
  461.       type: int64 default: 0

cache_bench

  1. ./cache_bench --help
  2. cache_bench: Warning: SetUsageMessage() never called
  3. ...
  4.   Flags from cache/cache_bench.cc:
  5.     -cache_size (Number of bytes to use as a cache of uncompressed data.)
  6.       type: int64 default: 8388608
  7.     -erase_percent (Ratio of erase to total workload (expressed as a
  8.       percentage)) type: int32 default: 10
  9.     -insert_percent (Ratio of insert to total workload (expressed as a
  10.       percentage)) type: int32 default: 40
  11.     -lookup_percent (Ratio of lookup to total workload (expressed as a
  12.       percentage)) type: int32 default: 50
  13.     -max_key (Max number of key to place in cache) type: int64
  14.       default: 1073741824
  15.     -num_shard_bits (shard_bits.) type: int32 default: 4
  16.     -ops_per_thread (Number of operations per thread.) type: uint64
  17.       default: 1200000
  18.     -populate_cache (Populate cache before operations) type: bool
  19.       default: false
  20.     -threads (Number of concurrent threads to run.) type: int32 default: 16
  21.     -use_clock_cache () type: bool default: false

persistent_cache_bench

  1. $ ./persistent_cache_bench -help
  2. persistent_cache_bench:
  3. USAGE:
  4. ./persistent_cache_bench [OPTIONS]...
  5. ...
  6.   Flags from utilities/persistent_cache/persistent_cache_bench.cc:
  7.     -benchmark (Benchmark mode) type: bool default: false
  8.     -cache_size (Cache size) type: uint64 default: 18446744073709551615
  9.     -cache_type (Cache type. (block_cache, volatile, tiered)) type: string
  10.       default: "block_cache"
  11.     -enable_pipelined_writes (Enable async writes) type: bool default: false
  12.     -iosize (Read IO size) type: int32 default: 4096
  13.     -log_path (Path for the log file) type: string default: "/tmp/log"
  14.     -nsec (nsec) type: int32 default: 10
  15.     -nthread_read (Lookup threads) type: int32 default: 1
  16.     -nthread_write (Insert threads) type: int32 default: 1
  17.     -path (Path for cachefile) type: string default: "/tmp/microbench/blkcache"
  18.     -volatile_cache_pct (Percentage of cache in memory tier.) type: int32
  19.       default: 10
  20.     -writer_iosize (File writer IO size) type: int32 default: 4096
  21.     -writer_qdepth (File writer qdepth) type: int32 default: 1