The difference is basically noise per the usual YCSB evaluation. Small
differences in workloads D and F (slightly worse) and workload E (slightly
better) that do not indicate serious regression.
Linux version 4.14.55-62.37.amzn1.x86_64
c3.8xlarge x 5
OpenJDK Runtime Environment (build 1.8.0_181-shenandoah-b13)
-Xms20g -Xmx20g -XX:+UseG1GC -XX:+AlwaysPreTouch -XX:+UseNUMA
-XX:-UseBiasedLocking -XX:+ParallelRefProcEnabled
Hadoop 2.9.2
Init: Load 100 M rows and snapshot
Run: Delete table, clone and redeploy from snapshot, run 10 M operations
Args: -threads 100 -target 50000
Test table: {NAME => 'u', BLOOMFILTER => 'ROW', VERSIONS => '1', IN_MEMORY
=> 'false', KEEP_DELETED_CELLS => 'FALSE', DATA_BLOCK_ENCODING =>
'ROW_INDEX_V1', TTL => 'FOREVER', COMPRESSION => 'SNAPPY', MIN_VERSIONS =>
'0', BLOCKCACHE => 'true', BLOCKSIZE => '65536', REPLICATION_SCOPE =>
'0'}
YCSB Workload A
target 50k/op/s 1.4.9 1.5.0
[OVERALL], RunTime(ms) 200592 200583
[OVERALL], Throughput(ops/sec) 49852 49855
[READ], AverageLatency(us) 544 559
[READ], MinLatency(us) 267 292
[READ], MaxLatency(us) 165631 185087
[READ], 95thPercentileLatency(us) 738 742
[READ], 99thPercentileLatency(us), 1877 1961
[UPDATE], AverageLatency(us) 1370 1181
[UPDATE], MinLatency(us) 702 646
[UPDATE], MaxLatency(us) 180735 177279
[UPDATE], 95thPercentileLatency(us) 1943 1652
[UPDATE], 99thPercentileLatency(us) 3257 3085
YCSB Workload B
target 50k/op/s 1.4.9 1.5.0
[OVERALL], RunTime(ms) 200599 200581
[OVERALL], Throughput(ops/sec) 49850 49855
[READ], AverageLatency(us), 454 471
[READ], MinLatency(us) 203 213
[READ], MaxLatency(us) 183423 174207
[READ], 95thPercentileLatency(us) 563 599
[READ], 99thPercentileLatency(us) 1360 1172
[UPDATE], AverageLatency(us) 1064 1029
[UPDATE], MinLatency(us) 746 726
[UPDATE], MaxLatency(us) 163455 101631
[UPDATE], 95thPercentileLatency(us) 1327 1157
[UPDATE], 99thPercentileLatency(us) 2241 1898
YCSB Workload C
target 50k/op/s 1.4.9 1.5.0
[OVERALL], RunTime(ms) 200541 200538
[OVERALL], Throughput(ops/sec) 49865 49865
[READ], AverageLatency(us) 332 327
[READ], MinLatency(us) 175 179
[READ], MaxLatency(us) 210559 170367
[READ], 95thPercentileLatency(us) 410 396
[READ], 99thPercentileLatency(us) 871 892
YCSB Workload D
target 50k/op/s 1.4.9 1.5.0
[OVERALL], RunTime(ms) 200579 200562
[OVERALL], Throughput(ops/sec) 49855 49859
[READ], AverageLatency(us) 487 547
[READ], MinLatency(us) 210 214
[READ], MaxLatency(us) 192255 177535
[READ], 95thPercentileLatency(us) 973 1529
[READ], 99thPercentileLatency(us) 1836 2683
[INSERT], AverageLatency(us) 1239 1152
[INSERT], MinLatency(us) 807 788
[INSERT], MaxLatency(us) 184575 148735
[INSERT], 95thPercentileLatency(us) 1496 1243
[INSERT], 99thPercentileLatency(us) 2965 2495
YCSB Workload E
target 10k/op/s 1.4.9 1.5.0
[OVERALL], RunTime(ms) 100605 100568
[OVERALL], Throughput(ops/sec) 9939 9943
[SCAN], AverageLatency(us) 3548 2687
[SCAN], MinLatency(us) 696 678
[SCAN], MaxLatency(us) 1059839 238463
[SCAN], 95thPercentileLatency(us) 8327 6791
[SCAN], 99thPercentileLatency(us) 17647 14415
[INSERT], AverageLatency(us) 2688 1555
[INSERT], MinLatency(us) 887 815
[INSERT], MaxLatency(us) 173311 154623
[INSERT], 95thPercentileLatency(us) 4455 2571
[INSERT], 99thPercentileLatency(us) 9303 5375
YCSB Workload F
target 50k/op/s 1.4.9 1.5.0
[OVERALL], RunTime(ms) 200562 204178
[OVERALL], Throughput(ops/sec) 49859 48976
[READ], AverageLatency(us) 856 1137
[READ], MinLatency(us) 262 257
[READ], MaxLatency(us) 205567 222335
[READ], 95thPercentileLatency(us) 2365 3475
[READ], 99thPercentileLatency(us) 3099 4143
[READ-MODIFY-WRITE], AverageLatency(us) 2559 2917
[READ-MODIFY-WRITE], MinLatency(us) 1100 1034
[READ-MODIFY-WRITE], MaxLatency(us) 208767 204799
[READ-MODIFY-WRITE], 95thPercentileLatency(us) 5747 7627
[READ-MODIFY-WRITE], 99thPercentileLatency(us) 7203 8919
[UPDATE], AverageLatency(us) 1700 1777
[UPDATE], MinLatency(us) 737 687
[UPDATE], MaxLatency(us) 97983 94271
[UPDATE], 95thPercentileLatency(us) 3377 4147
[UPDATE], 99thPercentileLatency(us) 4147 4831
On Thu, Apr 4, 2019 at 1:14 AM Yu Li <[email protected]> wrote:
> Thanks for the efforts boss.
>
> Since it's a new minor release, do we have performance comparison report
> with 1.4.9 as we did when releasing 1.4.0? If so, any reference? Many
> thanks!
>
> Best Regards,
> Yu
>
>
> On Thu, 4 Apr 2019 at 07:44, Andrew Purtell <[email protected]> wrote:
>
> > The fourth HBase 1.5.0 release candidate (RC3) is available for download
> at
> > https://dist.apache.org/repos/dist/dev/hbase/hbase-1.5.0RC3/ and Maven
> > artifacts are available in the temporary repository
> > https://repository.apache.org/content/repositories/orgapachehbase-1292/
> >
> > The git tag corresponding to the candidate is '1.5.0RC3’ (b0bc7225c5).
> >
> > A detailed source and binary compatibility report for this release is
> > available for your review at
> >
> >
> https://dist.apache.org/repos/dist/dev/hbase/hbase-1.5.0RC3/compat-check-report.html
> > .
> >
> > A list of the 115 issues resolved in this release can be found at
> > https://s.apache.org/K4Wk . The 1.5.0 changelog is derived from the
> > changelog of the last branch-1.4 release, 1.4.9.
> >
> > Please try out the candidate and vote +1/0/-1.
> >
> > The vote will be open for at least 72 hours. Unless objection I will try
> to
> > close it Friday April 12, 2019 if we have sufficient votes.
> >
> > Prior to making this announcement I made the following preflight checks:
> >
> > RAT check passes (7u80)
> > Unit test suite passes (7u80, 8u181)*
> > Opened the UI in a browser, poked around
> > LTT load 100M rows with 100% verification and 20% updates (8u181)
> > ITBLL 1B rows with slowDeterministic monkey (8u181)
> > ITBLL 1B rows with serverKilling monkey (8u181)
> >
> > There are known flaky tests. See HBASE-21904 and HBASE-21905. These flaky
> > tests do not represent serious test failures that would prevent a
> release.
> >
> >
> > --
> > Best regards,
> > Andrew
> >
>
--
Best regards,
Andrew
Words like orphans lost among the crosstalk, meaning torn from truth's
decrepit hands
- A23, Crosstalk
