NextGraph
/
rocksdb
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Several small improvements (#10803)

Browse Source 
Summary:
This has several small improvements.

benchmark.sh
* add BYTES_PER_SYNC as an env variable
* use --prepopulate_block_cache when O_DIRECT is used
* use --undefok to list options that don't work for all 7.x releases
* print "failure" in report.tsv when a benchmark fails
* parse the slightly different throughput line used by db_bench for multireadrandom
* remove the trailing comma for BlobDB size before printing it in report.tsv
* use the last line of the output from /bin/time as there can be more than one line when db_bench has a non-zero exit
* fix more bash lint warnings
* add ",stats" to the --benchmark=... lines to get stats at the end of each benchmark

benchmark_compare.sh
* run revrange immediately after fillseq to let compaction debt get removed
* add --multiread_batched when --benchmarks=multireadrandom is used
* use --benchmarks=overwriteandwait when supported to get a more accurate measure of write-amp

Pull Request resolved: https://github.com/facebook/rocksdb/pull/10803

Test Plan: Run it for leveled, universal and BlobDB

Reviewed By: jay-zhuang

Differential Revision: D40278315

Pulled By: mdcallag

fbshipit-source-id: 793134ddc7d48d05a07436cd8942c375a23983a7
main
Mark Callaghan 2 years ago committed by Facebook GitHub Bot
parent 23b7dc2f4f
commit 6ff0c204cb
2 changed files with 104 additions and 67 deletions
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Show Stats Download Patch File Download Diff File
  1. 149
      tools/benchmark.sh
  2. 22
      tools/benchmark_compare.sh

149
tools/benchmark.sh
Unescape View File

@ -67,6 +67,7 @@ function display_usage() {
echo -e "\tMAX_BACKGROUND_JOBS\t\tThe value for max_background_jobs (default: 16)" echo -e "\tMAX_BACKGROUND_JOBS\t\tThe value for max_background_jobs (default: 16)"
echo -e "\tCACHE_INDEX_AND_FILTER_BLOCKS\tThe value for cache_index_and_filter_blocks (default: 0)" echo -e "\tCACHE_INDEX_AND_FILTER_BLOCKS\tThe value for cache_index_and_filter_blocks (default: 0)"
echo -e "\tUSE_O_DIRECT\t\t\tUse O_DIRECT for user reads and compaction" echo -e "\tUSE_O_DIRECT\t\t\tUse O_DIRECT for user reads and compaction"
echo -e "\tBYTES_PER_SYNC\t\t\tValue for bytes_per_sync, set to zero when USE_O_DIRECT is true"
echo -e "\tSTATS_INTERVAL_SECONDS\t\tValue for stats_interval_seconds" echo -e "\tSTATS_INTERVAL_SECONDS\t\tValue for stats_interval_seconds"
echo -e "\tREPORT_INTERVAL_SECONDS\t\tValue for report_interval_seconds" echo -e "\tREPORT_INTERVAL_SECONDS\t\tValue for report_interval_seconds"
echo -e "\tSUBCOMPACTIONS\t\t\tValue for subcompactions" echo -e "\tSUBCOMPACTIONS\t\t\tValue for subcompactions"
@ -182,7 +183,7 @@ target_file_mb=${TARGET_FILE_SIZE_BASE_MB:-128}
l1_mb=${MAX_BYTES_FOR_LEVEL_BASE_MB:-1024} l1_mb=${MAX_BYTES_FOR_LEVEL_BASE_MB:-1024}
max_background_jobs=${MAX_BACKGROUND_JOBS:-16} max_background_jobs=${MAX_BACKGROUND_JOBS:-16}
stats_interval_seconds=${STATS_INTERVAL_SECONDS:-60} stats_interval_seconds=${STATS_INTERVAL_SECONDS:-60}
report_interval_seconds=${REPORT_INTERVAL_SECONDS:-5} report_interval_seconds=${REPORT_INTERVAL_SECONDS:-1}
subcompactions=${SUBCOMPACTIONS:-1} subcompactions=${SUBCOMPACTIONS:-1}
per_level_fanout=${PER_LEVEL_FANOUT:-8} per_level_fanout=${PER_LEVEL_FANOUT:-8}
@ -214,9 +215,11 @@ fi
o_direct_flags="" o_direct_flags=""
if [ ! -z $USE_O_DIRECT ]; then if [ ! -z $USE_O_DIRECT ]; then
# TODO: deal with flags only supported in new versions, like prepopulate_block_cache # Some of these flags are only supported in new versions and --undefok makes that work
#o_direct_flags="--use_direct_reads --use_direct_io_for_flush_and_compaction --prepopulate_block_cache=1" o_direct_flags="--use_direct_reads --use_direct_io_for_flush_and_compaction --prepopulate_block_cache=1"
o_direct_flags="--use_direct_reads --use_direct_io_for_flush_and_compaction" bytes_per_sync=0
else
bytes_per_sync=${BYTES_PER_SYNC:-$(( 1 * M ))}
fi fi
univ_min_merge_width=${UNIVERSAL_MIN_MERGE_WIDTH:-2} univ_min_merge_width=${UNIVERSAL_MIN_MERGE_WIDTH:-2}
@ -242,7 +245,18 @@ blob_cache_size=${BLOB_CACHE_SIZE:-$(( 16 * $G ))}
blob_cache_numshardbits=${BLOB_CACHE_NUMSHARDBITS:-6} blob_cache_numshardbits=${BLOB_CACHE_NUMSHARDBITS:-6}
prepopulate_blob_cache=${PREPOPULATE_BLOB_CACHE:-0} prepopulate_blob_cache=${PREPOPULATE_BLOB_CACHE:-0}
# This script still works back to RocksDB 6.0
undef_params="\
use_blob_cache,\
use_shared_block_and_blob_cache,\
blob_cache_size,blob_cache_numshardbits,\
prepopulate_blob_cache,\
multiread_batched,\
cache_low_pri_pool_ratio,\
prepopulate_block_cache"
const_params_base=" const_params_base="
--undefok=$undef_params \
--db=$DB_DIR \ --db=$DB_DIR \
--wal_dir=$WAL_DIR \ --wal_dir=$WAL_DIR \
\ \
@ -255,7 +269,7 @@ const_params_base="
--compression_max_dict_bytes=$compression_max_dict_bytes \ --compression_max_dict_bytes=$compression_max_dict_bytes \
--compression_ratio=0.5 \ --compression_ratio=0.5 \
--compression_type=$compression_type \ --compression_type=$compression_type \
--bytes_per_sync=$((8 * M)) \ --bytes_per_sync=$bytes_per_sync \
$cache_meta_flags \ $cache_meta_flags \
$o_direct_flags \ $o_direct_flags \
--benchmark_write_rate_limit=$(( 1024 * 1024 * $mb_written_per_sec )) \ --benchmark_write_rate_limit=$(( 1024 * 1024 * $mb_written_per_sec )) \
@ -293,7 +307,6 @@ level_const_params="
" "
# These inherit level_const_params because the non-blob LSM tree uses leveled compaction. # These inherit level_const_params because the non-blob LSM tree uses leveled compaction.
# The use of undefok is for options that are not supported until 7.5.
blob_const_params=" blob_const_params="
$level_const_params \ $level_const_params \
--enable_blob_files=true \ --enable_blob_files=true \
@ -309,7 +322,6 @@ blob_const_params="
--blob_cache_size=$blob_cache_size \ --blob_cache_size=$blob_cache_size \
--blob_cache_numshardbits=$blob_cache_numshardbits \ --blob_cache_numshardbits=$blob_cache_numshardbits \
--prepopulate_blob_cache=$prepopulate_blob_cache \ --prepopulate_blob_cache=$prepopulate_blob_cache \
--undefok=use_blob_cache,use_shared_block_and_blob_cache,blob_cache_size,blob_cache_numshardbits,prepopulate_blob_cache \
" "
# TODO: # TODO:
@ -511,20 +523,32 @@ function summarize_result {
stall_pct=$( grep "^Cumulative stall" $test_out| tail -1 | awk '{ print $5 }' ) stall_pct=$( grep "^Cumulative stall" $test_out| tail -1 | awk '{ print $5 }' )
nstall=$( grep ^Stalls\(count\): $test_out | tail -1 | awk '{ print $2 + $6 + $10 + $14 + $18 + $20 }' ) nstall=$( grep ^Stalls\(count\): $test_out | tail -1 | awk '{ print $2 + $6 + $10 + $14 + $18 + $20 }' )
if ! grep ^"$bench_name" "$test_out" > /dev/null 2>&1 ; then
echo -e "failed\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t$test_name\t$my_date\t$version\t$job_id\t$git_hash"
return
fi
# Output formats # Output formats
# V1: readrandom : 10.218 micros/op 3131616 ops/sec; 1254.3 MB/s (176144999 of 176144999 found)
# The MB/s is mssing for multireadrandom
# V1a: multireadrandom : 10.164 micros/op 3148272 ops/sec; (177099990 of 177099990 found)
# V1: overwrite : 7.939 micros/op 125963 ops/sec; 50.5 MB/s # V1: overwrite : 7.939 micros/op 125963 ops/sec; 50.5 MB/s
# V2: overwrite : 7.854 micros/op 127320 ops/sec 1800.001 seconds 229176999 operations; 51.0 MB/s # V2: overwrite : 7.854 micros/op 127320 ops/sec 1800.001 seconds 229176999 operations; 51.0 MB/s
format_version=$( grep ^${bench_name} $test_out \ format_version=$( grep ^"$bench_name" "$test_out" \
| awk '{ if (NF >= 10 && $8 == "seconds") { print "V2" } else { print "V1" } }' ) | awk '{ if (NF >= 10 && $8 == "seconds") { print "V2" } else { print "V1" } }' )
if [ $format_version == "V1" ]; then if [ $format_version == "V1" ]; then
ops_sec=$( grep ^${bench_name} $test_out | awk '{ print $5 }' ) ops_sec=$( grep ^"$bench_name" "$test_out" | awk '{ print $5 }' )
usecs_op=$( grep ^${bench_name} $test_out | awk '{ printf "%.1f", $3 }' ) usecs_op=$( grep ^"$bench_name" "$test_out" | awk '{ printf "%.1f", $3 }' )
mb_sec=$( grep ^${bench_name} $test_out | awk '{ print $7 }' ) if [ "$bench_name" == "multireadrandom" ]; then
mb_sec="NA"
else
mb_sec=$( grep ^"$bench_name" "$test_out" | awk '{ print $7 }' )
fi
else else
ops_sec=$( grep ^${bench_name} $test_out | awk '{ print $5 }' ) ops_sec=$( grep ^"$bench_name" "$test_out" | awk '{ print $5 }' )
usecs_op=$( grep ^${bench_name} $test_out | awk '{ printf "%.1f", $3 }' ) usecs_op=$( grep ^"$bench_name" "$test_out" | awk '{ printf "%.1f", $3 }' )
mb_sec=$( grep ^${bench_name} $test_out | awk '{ print $11 }' ) mb_sec=$( grep ^"$bench_name" "$test_out" | awk '{ print $11 }' )
fi fi
# For RocksDB version 4.x there are fewer fields but this still parses correctly # For RocksDB version 4.x there are fewer fields but this still parses correctly
@ -555,8 +579,10 @@ function summarize_result {
c_wsecs=$( grep "^ Sum" $test_out | tail -1 | awk '{ printf "%.0f", $15 }' ) c_wsecs=$( grep "^ Sum" $test_out | tail -1 | awk '{ printf "%.0f", $15 }' )
c_csecs=$( grep "^ Sum" $test_out | tail -1 | awk '{ printf "%.0f", $16 }' ) c_csecs=$( grep "^ Sum" $test_out | tail -1 | awk '{ printf "%.0f", $16 }' )
lsm_size=$( grep "^ Sum" $test_out | tail -1 | awk '{ printf "%.0f%s", $3, $4 }' ) lsm_size=$( grep "^ Sum" "$test_out" | tail -1 | awk '{ printf "%.0f%s", $3, $4 }' )
blob_size=$( grep "^Blob file count:" $test_out | tail -1 | awk '{ printf "%s%s", $7, $8 }' ) blob_size=$( grep "^Blob file count:" "$test_out" | tail -1 | awk '{ printf "%.0f%s", $7, $8 }' )
# Remove the trailing comma from blob_size: 3.0GB, -> 3.0GB
blob_size="${blob_size/,/}"
b_rgb=$( grep "^ Sum" $test_out | tail -1 | awk '{ printf "%.0f", $21 }' ) b_rgb=$( grep "^ Sum" $test_out | tail -1 | awk '{ printf "%.0f", $21 }' )
b_wgb=$( grep "^ Sum" $test_out | tail -1 | awk '{ printf "%.0f", $22 }' ) b_wgb=$( grep "^ Sum" $test_out | tail -1 | awk '{ printf "%.0f", $22 }' )
@ -567,47 +593,48 @@ function summarize_result {
p9999=$( grep "^Percentiles:" $test_out | tail -1 | awk '{ printf "%.0f", $11 }' ) p9999=$( grep "^Percentiles:" $test_out | tail -1 | awk '{ printf "%.0f", $11 }' )
pmax=$( grep "^Min: " $test_out | grep Median: | grep Max: | awk '{ printf "%.0f", $6 }' ) pmax=$( grep "^Min: " $test_out | grep Median: | grep Max: | awk '{ printf "%.0f", $6 }' )
time_out=$test_out.time # Use the last line because there might be extra lines when the db_bench process exits with an error
u_cpu=$( awk '{ printf "%.1f", $2 / 1000.0 }' $time_out ) time_out="$test_out".time
s_cpu=$( awk '{ printf "%.1f", $3 / 1000.0 }' $time_out ) u_cpu=$( tail -1 "$time_out" | awk '{ printf "%.1f", $2 / 1000.0 }' )
s_cpu=$( tail -1 "$time_out" | awk '{ printf "%.1f", $3 / 1000.0 }' )
rss="NA" rss="NA"
if [ -f $test_out.stats.ps ]; then if [ -f $test_out.stats.ps ]; then
rss=$( tail -1 $test_out.stats.ps | awk '{ printf "%.1f\n", $6 / (1024 * 1024) }' ) rss=$( awk '{ printf "%.1f\n", $6 / (1024 * 1024) }' "$test_out".stats.ps | sort -n | tail -1 )
fi fi
# if the report TSV (Tab Separate Values) file does not yet exist, create it and write the header row to it # if the report TSV (Tab Separate Values) file does not yet exist, create it and write the header row to it
if [ ! -f "$report" ]; then if [ ! -f "$report" ]; then
echo -e "# ops_sec - operations per second" >> $report echo -e "# ops_sec - operations per second" >> "$report"
echo -e "# mb_sec - ops_sec * size-of-operation-in-MB" >> $report echo -e "# mb_sec - ops_sec * size-of-operation-in-MB" >> "$report"
echo -e "# lsm_sz - size of LSM tree" >> $report echo -e "# lsm_sz - size of LSM tree" >> "$report"
echo -e "# blob_sz - size of BlobDB logs" >> $report echo -e "# blob_sz - size of BlobDB logs" >> "$report"
echo -e "# c_wgb - GB written by compaction" >> $report echo -e "# c_wgb - GB written by compaction" >> "$report"
echo -e "# w_amp - Write-amplification as (bytes written by compaction / bytes written by memtable flush)" >> $report echo -e "# w_amp - Write-amplification as (bytes written by compaction / bytes written by memtable flush)" >> "$report"
echo -e "# c_mbps - Average write rate for compaction" >> $report echo -e "# c_mbps - Average write rate for compaction" >> "$report"
echo -e "# c_wsecs - Wall clock seconds doing compaction" >> $report echo -e "# c_wsecs - Wall clock seconds doing compaction" >> "$report"
echo -e "# c_csecs - CPU seconds doing compaction" >> $report echo -e "# c_csecs - CPU seconds doing compaction" >> "$report"
echo -e "# b_rgb - Blob compaction read GB" >> $report echo -e "# b_rgb - Blob compaction read GB" >> "$report"
echo -e "# b_wgb - Blob compaction write GB" >> $report echo -e "# b_wgb - Blob compaction write GB" >> "$report"
echo -e "# usec_op - Microseconds per operation" >> $report echo -e "# usec_op - Microseconds per operation" >> "$report"
echo -e "# p50, p99, p99.9, p99.99 - 50th, 99th, 99.9th, 99.99th percentile response time in usecs" >> $report echo -e "# p50, p99, p99.9, p99.99 - 50th, 99th, 99.9th, 99.99th percentile response time in usecs" >> "$report"
echo -e "# pmax - max response time in usecs" >> $report echo -e "# pmax - max response time in usecs" >> "$report"
echo -e "# uptime - RocksDB uptime in seconds" >> $report echo -e "# uptime - RocksDB uptime in seconds" >> "$report"
echo -e "# stall% - Percentage of time writes are stalled" >> $report echo -e "# stall% - Percentage of time writes are stalled" >> "$report"
echo -e "# Nstall - Number of stalls" >> $report echo -e "# Nstall - Number of stalls" >> "$report"
echo -e "# u_cpu - #seconds/1000 of user CPU" >> $report echo -e "# u_cpu - #seconds/1000 of user CPU" >> "$report"
echo -e "# s_cpu - #seconds/1000 of system CPU" >> $report echo -e "# s_cpu - #seconds/1000 of system CPU" >> "$report"
echo -e "# rss - max RSS in GB for db_bench process" >> $report echo -e "# rss - max RSS in GB for db_bench process" >> "$report"
echo -e "# test - Name of test" >> $report echo -e "# test - Name of test" >> "$report"
echo -e "# date - Date/time of test" >> $report echo -e "# date - Date/time of test" >> "$report"
echo -e "# version - RocksDB version" >> $report echo -e "# version - RocksDB version" >> "$report"
echo -e "# job_id - User-provided job ID" >> $report echo -e "# job_id - User-provided job ID" >> "$report"
echo -e "# githash - git hash at which db_bench was compiled" echo -e "# githash - git hash at which db_bench was compiled" >> "$report"
echo -e $tsv_header >> $report echo -e $tsv_header >> "$report"
fi fi
echo -e "$ops_sec\t$mb_sec\t$lsm_size\t$blob_size\t$sum_wgb\t$wamp\t$cmb_ps\t$c_wsecs\t$c_csecs\t$b_rgb\t$b_wgb\t$usecs_op\t$p50\t$p99\t$p999\t$p9999\t$pmax\t$uptime\t$stall_pct\t$nstall\t$u_cpu\t$s_cpu\t$rss\t$test_name\t$my_date\t$version\t$job_id\t$git_hash" \ echo -e "$ops_sec\t$mb_sec\t$lsm_size\t$blob_size\t$sum_wgb\t$wamp\t$cmb_ps\t$c_wsecs\t$c_csecs\t$b_rgb\t$b_wgb\t$usecs_op\t$p50\t$p99\t$p999\t$p9999\t$pmax\t$uptime\t$stall_pct\t$nstall\t$u_cpu\t$s_cpu\t$rss\t$test_name\t$my_date\t$version\t$job_id\t$git_hash" \
>> $report >> "$report"
} }
function run_bulkload { function run_bulkload {
@ -616,7 +643,7 @@ function run_bulkload {
echo "Bulk loading $num_keys random keys" echo "Bulk loading $num_keys random keys"
log_file_name=$output_dir/benchmark_bulkload_fillrandom.log log_file_name=$output_dir/benchmark_bulkload_fillrandom.log
time_cmd=$( get_cmd $log_file_name.time ) time_cmd=$( get_cmd $log_file_name.time )
cmd="$time_cmd ./db_bench --benchmarks=fillrandom \ cmd="$time_cmd ./db_bench --benchmarks=fillrandom,stats \
--use_existing_db=0 \ --use_existing_db=0 \
--disable_auto_compactions=1 \ --disable_auto_compactions=1 \
--sync=0 \ --sync=0 \
@ -640,7 +667,7 @@ function run_bulkload {
echo "Compacting..." echo "Compacting..."
log_file_name=$output_dir/benchmark_bulkload_compact.log log_file_name=$output_dir/benchmark_bulkload_compact.log
time_cmd=$( get_cmd $log_file_name.time ) time_cmd=$( get_cmd $log_file_name.time )
cmd="$time_cmd ./db_bench --benchmarks=compact \ cmd="$time_cmd ./db_bench --benchmarks=compact,stats \
--use_existing_db=1 \ --use_existing_db=1 \
--disable_auto_compactions=1 \ --disable_auto_compactions=1 \
--sync=0 \ --sync=0 \
@ -680,7 +707,7 @@ function run_manual_compaction_worker {
# Make sure that fillrandom uses the same compaction options as compact. # Make sure that fillrandom uses the same compaction options as compact.
time_cmd=$( get_cmd $log_file_name.time ) time_cmd=$( get_cmd $log_file_name.time )
cmd="$time_cmd ./db_bench --benchmarks=fillrandom \ cmd="$time_cmd ./db_bench --benchmarks=fillrandom,stats \
--use_existing_db=0 \ --use_existing_db=0 \
--disable_auto_compactions=0 \ --disable_auto_compactions=0 \
--sync=0 \ --sync=0 \
@ -714,7 +741,7 @@ function run_manual_compaction_worker {
# doesn't output regular statistics then we'll just use the time command to # doesn't output regular statistics then we'll just use the time command to
# measure how long this step takes. # measure how long this step takes.
cmd="{ \ cmd="{ \
time ./db_bench --benchmarks=compact \ time ./db_bench --benchmarks=compact,stats \
--use_existing_db=1 \ --use_existing_db=1 \
--disable_auto_compactions=0 \ --disable_auto_compactions=0 \
--sync=0 \ --sync=0 \
@ -800,7 +827,7 @@ function run_fillseq {
echo "Loading $num_keys keys sequentially" echo "Loading $num_keys keys sequentially"
time_cmd=$( get_cmd $log_file_name.time ) time_cmd=$( get_cmd $log_file_name.time )
cmd="$time_cmd ./db_bench --benchmarks=fillseq \ cmd="$time_cmd ./db_bench --benchmarks=fillseq,stats \
$params_fillseq \ $params_fillseq \
$comp_arg \ $comp_arg \
--use_existing_db=0 \ --use_existing_db=0 \
@ -843,7 +870,7 @@ function run_lsm {
log_file_name=$output_dir/benchmark_${job}.log log_file_name=$output_dir/benchmark_${job}.log
time_cmd=$( get_cmd $log_file_name.time ) time_cmd=$( get_cmd $log_file_name.time )
cmd="$time_cmd ./db_bench --benchmarks=$benchmarks \ cmd="$time_cmd ./db_bench --benchmarks=$benchmarks,stats \
--use_existing_db=1 \ --use_existing_db=1 \
--sync=0 \ --sync=0 \
$params_w \ $params_w \
@ -871,7 +898,7 @@ function run_change {
echo "Do $num_keys random $output_name" echo "Do $num_keys random $output_name"
log_file_name="$output_dir/benchmark_${output_name}.t${num_threads}.s${syncval}.log" log_file_name="$output_dir/benchmark_${output_name}.t${num_threads}.s${syncval}.log"
time_cmd=$( get_cmd $log_file_name.time ) time_cmd=$( get_cmd $log_file_name.time )
cmd="$time_cmd ./db_bench --benchmarks=$benchmarks \ cmd="$time_cmd ./db_bench --benchmarks=$benchmarks,stats \
--use_existing_db=1 \ --use_existing_db=1 \
--sync=$syncval \ --sync=$syncval \
$params_w \ $params_w \
@ -896,7 +923,7 @@ function run_filluniquerandom {
echo "Loading $num_keys unique keys randomly" echo "Loading $num_keys unique keys randomly"
log_file_name=$output_dir/benchmark_filluniquerandom.log log_file_name=$output_dir/benchmark_filluniquerandom.log
time_cmd=$( get_cmd $log_file_name.time ) time_cmd=$( get_cmd $log_file_name.time )
cmd="$time_cmd ./db_bench --benchmarks=filluniquerandom \ cmd="$time_cmd ./db_bench --benchmarks=filluniquerandom,stats \
--use_existing_db=0 \ --use_existing_db=0 \
--sync=0 \ --sync=0 \
$params_w \ $params_w \
@ -920,7 +947,7 @@ function run_readrandom {
echo "Reading $num_keys random keys" echo "Reading $num_keys random keys"
log_file_name="${output_dir}/benchmark_readrandom.t${num_threads}.log" log_file_name="${output_dir}/benchmark_readrandom.t${num_threads}.log"
time_cmd=$( get_cmd $log_file_name.time ) time_cmd=$( get_cmd $log_file_name.time )
cmd="$time_cmd ./db_bench --benchmarks=readrandom \ cmd="$time_cmd ./db_bench --benchmarks=readrandom,stats \
--use_existing_db=1 \ --use_existing_db=1 \
$params_w \ $params_w \
--threads=$num_threads \ --threads=$num_threads \
@ -943,7 +970,7 @@ function run_multireadrandom {
echo "Multi-Reading $num_keys random keys" echo "Multi-Reading $num_keys random keys"
log_file_name="${output_dir}/benchmark_multireadrandom.t${num_threads}.log" log_file_name="${output_dir}/benchmark_multireadrandom.t${num_threads}.log"
time_cmd=$( get_cmd $log_file_name.time ) time_cmd=$( get_cmd $log_file_name.time )
cmd="$time_cmd ./db_bench --benchmarks=multireadrandom \ cmd="$time_cmd ./db_bench --benchmarks=multireadrandom,stats \
--use_existing_db=1 \ --use_existing_db=1 \
--threads=$num_threads \ --threads=$num_threads \
--batch_size=10 \ --batch_size=10 \
@ -968,7 +995,7 @@ function run_readwhile {
echo "Reading $num_keys random keys while $operation" echo "Reading $num_keys random keys while $operation"
log_file_name="${output_dir}/benchmark_readwhile${operation}.t${num_threads}.log" log_file_name="${output_dir}/benchmark_readwhile${operation}.t${num_threads}.log"
time_cmd=$( get_cmd $log_file_name.time ) time_cmd=$( get_cmd $log_file_name.time )
cmd="$time_cmd ./db_bench --benchmarks=readwhile${operation} \ cmd="$time_cmd ./db_bench --benchmarks=readwhile${operation},stats \
--use_existing_db=1 \ --use_existing_db=1 \
--sync=$syncval \ --sync=$syncval \
$params_w \ $params_w \
@ -996,7 +1023,7 @@ function run_rangewhile {
log_file_name="${output_dir}/benchmark_${full_name}.t${num_threads}.log" log_file_name="${output_dir}/benchmark_${full_name}.t${num_threads}.log"
time_cmd=$( get_cmd $log_file_name.time ) time_cmd=$( get_cmd $log_file_name.time )
echo "Range scan $num_keys random keys while ${operation} for reverse_iter=${reverse_arg}" echo "Range scan $num_keys random keys while ${operation} for reverse_iter=${reverse_arg}"
cmd="$time_cmd ./db_bench --benchmarks=seekrandomwhile${operation} \ cmd="$time_cmd ./db_bench --benchmarks=seekrandomwhile${operation},stats \
--use_existing_db=1 \ --use_existing_db=1 \
--sync=$syncval \ --sync=$syncval \
$params_w \ $params_w \
@ -1020,7 +1047,7 @@ function run_range {
log_file_name="${output_dir}/benchmark_${full_name}.t${num_threads}.log" log_file_name="${output_dir}/benchmark_${full_name}.t${num_threads}.log"
time_cmd=$( get_cmd $log_file_name.time ) time_cmd=$( get_cmd $log_file_name.time )
echo "Range scan $num_keys random keys for reverse_iter=${reverse_arg}" echo "Range scan $num_keys random keys for reverse_iter=${reverse_arg}"
cmd="$time_cmd ./db_bench --benchmarks=seekrandom \ cmd="$time_cmd ./db_bench --benchmarks=seekrandom,stats \
--use_existing_db=1 \ --use_existing_db=1 \
$params_w \ $params_w \
--threads=$num_threads \ --threads=$num_threads \
@ -1045,7 +1072,7 @@ function run_randomtransaction {
echo "..." echo "..."
log_file_name=$output_dir/benchmark_randomtransaction.log log_file_name=$output_dir/benchmark_randomtransaction.log
time_cmd=$( get_cmd $log_file_name.time ) time_cmd=$( get_cmd $log_file_name.time )
cmd="$time_cmd ./db_bench $params_w --benchmarks=randomtransaction \ cmd="$time_cmd ./db_bench $params_w --benchmarks=randomtransaction,stats \
--num=$num_keys \ --num=$num_keys \
--transaction_db \ --transaction_db \
--threads=5 \ --threads=5 \

22
tools/benchmark_compare.sh
Unescape View File

@ -32,6 +32,7 @@ max_background_jobs=${MAX_BACKGROUND_JOBS:-8}
stats_interval_seconds=${STATS_INTERVAL_SECONDS:-20} stats_interval_seconds=${STATS_INTERVAL_SECONDS:-20}
cache_index_and_filter_blocks=${CACHE_INDEX_AND_FILTER_BLOCKS:-0} cache_index_and_filter_blocks=${CACHE_INDEX_AND_FILTER_BLOCKS:-0}
# USE_O_DIRECT doesn't need a default # USE_O_DIRECT doesn't need a default
bytes_per_sync=${BYTES_PER_SYNC:-$(( 1 * M ))}
# CACHE_SIZE_MB doesn't need a default # CACHE_SIZE_MB doesn't need a default
min_level_to_compress=${MIN_LEVEL_TO_COMPRESS:-"-1"} min_level_to_compress=${MIN_LEVEL_TO_COMPRESS:-"-1"}
@ -86,6 +87,7 @@ base_args+=( MAX_BACKGROUND_JOBS="$max_background_jobs" )
base_args+=( STATS_INTERVAL_SECONDS="$stats_interval_seconds" ) base_args+=( STATS_INTERVAL_SECONDS="$stats_interval_seconds" )
base_args+=( CACHE_INDEX_AND_FILTER_BLOCKS="$cache_index_and_filter_blocks" ) base_args+=( CACHE_INDEX_AND_FILTER_BLOCKS="$cache_index_and_filter_blocks" )
base_args+=( COMPACTION_STYLE="$compaction_style" ) base_args+=( COMPACTION_STYLE="$compaction_style" )
base_args+=( BYTES_PER_SYNC="$bytes_per_sync" )
if [ -n "$USE_O_DIRECT" ]; then if [ -n "$USE_O_DIRECT" ]; then
base_args+=( USE_O_DIRECT=1 ) base_args+=( USE_O_DIRECT=1 )
@ -131,7 +133,7 @@ else
fi fi
function usage { function usage {
echo "usage: benchmark_wrapper.sh db_dir output_dir version+" echo "usage: benchmark_compare.sh db_dir output_dir version+"
echo -e "\tdb_dir\t\tcreate RocksDB database in this directory" echo -e "\tdb_dir\t\tcreate RocksDB database in this directory"
echo -e "\toutput_dir\twrite output from performance tests in this directory" echo -e "\toutput_dir\twrite output from performance tests in this directory"
echo -e "\tversion+\tspace separated sequence of RocksDB versions to test." echo -e "\tversion+\tspace separated sequence of RocksDB versions to test."
@ -155,6 +157,7 @@ function usage {
echo -e "\tMAX_BACKGROUND_JOBS\t\tvalue for max_background_jobs" echo -e "\tMAX_BACKGROUND_JOBS\t\tvalue for max_background_jobs"
echo -e "\tCACHE_INDEX_AND_FILTER_BLOCKS\tvalue for cache_index_and_filter_blocks" echo -e "\tCACHE_INDEX_AND_FILTER_BLOCKS\tvalue for cache_index_and_filter_blocks"
echo -e "\tUSE_O_DIRECT\t\t\tUse O_DIRECT for user reads and compaction" echo -e "\tUSE_O_DIRECT\t\t\tUse O_DIRECT for user reads and compaction"
echo -e "\tBYTES_PER_SYNC\t\t\tValue for bytes_per_sync"
echo -e "\tSTATS_INTERVAL_SECONDS\t\tvalue for stats_interval_seconds" echo -e "\tSTATS_INTERVAL_SECONDS\t\tvalue for stats_interval_seconds"
echo -e "\tSUBCOMPACTIONS\t\t\tvalue for subcompactions" echo -e "\tSUBCOMPACTIONS\t\t\tvalue for subcompactions"
echo -e "\tCOMPACTION_STYLE\t\tCompaction style to use, one of: leveled, universal, blob" echo -e "\tCOMPACTION_STYLE\t\tCompaction style to use, one of: leveled, universal, blob"
@ -258,19 +261,20 @@ for v in "$@" ; do
# Read-only tests. The LSM tree shape is in a deterministic state if trivial move # Read-only tests. The LSM tree shape is in a deterministic state if trivial move
# was used during the load. # was used during the load.
# Add revrange with a fixed duration and hardwired number of keys and threads to give
# compaction debt leftover from fillseq a chance at being removed. Not using waitforcompaction
# here because it isn't supported on older db_bench versions.
env -i "${args_nolim[@]}" DURATION=300 NUM_KEYS=100 NUM_THREADS=1 bash ./benchmark.sh revrange
env -i "${args_nolim[@]}" DURATION="$duration_ro" bash ./benchmark.sh readrandom env -i "${args_nolim[@]}" DURATION="$duration_ro" bash ./benchmark.sh readrandom
# Skipped for CI - a single essentail readrandom is enough to set up for other tests # Skipped for CI - a single essentail readrandom is enough to set up for other tests
if [ "$ci_tests_only" != "true" ]; then if [ "$ci_tests_only" != "true" ]; then
env -i "${args_nolim[@]}" DURATION="$duration_ro" bash ./benchmark.sh fwdrange env -i "${args_nolim[@]}" DURATION="$duration_ro" bash ./benchmark.sh fwdrange
env -i "${args_lim[@]}" DURATION="$duration_ro" bash ./benchmark.sh multireadrandom env -i "${args_lim[@]}" DURATION="$duration_ro" bash ./benchmark.sh multireadrandom --multiread_batched
else else
echo "CI_TESTS_ONLY is set, skipping optional read steps." echo "CI_TESTS_ONLY is set, skipping optional read steps."
fi fi
# Skipping --multiread_batched for now because it isn't supported on older 6.X releases
# env "${args_lim[@]}" DURATION=$duration_ro bash ./benchmark.sh multireadrandom --multiread_batched
# Write 10% of the keys. The goal is to randomize keys prior to Lmax # Write 10% of the keys. The goal is to randomize keys prior to Lmax
p10=$( echo "$num_keys" "$num_threads" | awk '{ printf "%.0f", $1 / $2 / 10.0 }' ) p10=$( echo "$num_keys" "$num_threads" | awk '{ printf "%.0f", $1 / $2 / 10.0 }' )
env -i "${args_nolim[@]}" WRITES="$p10" bash ./benchmark.sh overwritesome env -i "${args_nolim[@]}" WRITES="$p10" bash ./benchmark.sh overwritesome
@ -301,7 +305,13 @@ for v in "$@" ; do
# This creates much compaction debt which will be a problem for tests added after it. # This creates much compaction debt which will be a problem for tests added after it.
# Also, the compaction stats measured at test end can underestimate write-amp depending # Also, the compaction stats measured at test end can underestimate write-amp depending
# on how much compaction debt is allowed. # on how much compaction debt is allowed.
env -i "${args_nolim[@]}" DURATION="$duration_rw" bash ./benchmark.sh overwrite if [ "$compaction_style" == "leveled" ] && ./db_bench --benchmarks=waitforcompaction ; then
# Use waitforcompaction to get more accurate write-amp measurement
env -i "${args_nolim[@]}" DURATION="$duration_rw" bash ./benchmark.sh overwriteandwait
else
# waitforcompaction hangs with universal, see https://github.com/facebook/rocksdb/issues/9275
env -i "${args_nolim[@]}" DURATION="$duration_rw" bash ./benchmark.sh overwrite
fi
cp "$dbdir"/LOG* "$my_odir" cp "$dbdir"/LOG* "$my_odir"
gzip -9 "$my_odir"/LOG* gzip -9 "$my_odir"/LOG*

Write Preview
Loading…
Cancel
Save