hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Enis Söztutar <enis....@gmail.com>
Subject Re: Comparsion between 0.98.0 and 0.98.1RC1 using YCSB
Date Thu, 27 Mar 2014 22:35:40 GMT
BTW, this is not related, but we should not run with

dfs.client.read.shortcircuit.skip.checksum=true

This will cause WAL files read with short circuit to bypass checksum
verification.

We used to put up a warning in HBASE-8322, somehow it seems that part has
disappeared from trunk.

Enis

On Tue, Mar 25, 2014 at 12:29 AM, Andrew Purtell <apurtell@apache.org>wrote:

> Thought I'd look at how 0.98.0 and 0.98.1 RC1 compare using YCSB. It could
> be .1 is just a hair slower than .0, but the results are similar. For some
> workloads either .0 or .1 could be slightly ahead run to run. Results from
> the most recent runs are presented below.
>
> *Hardware and Versions*
>
>   Hadoop 2.2.0
>   HBase 0.98.1-hadoop2 RC1
>
>   5x EC2 cc2.8xlarge
>     32 cores
>     60 GB RAM
>     4 x 840 GB directly attached spinning media, as DN storage volumes
>     NameNode: 4 GB heap
>     DataNode: 1 GB heap
>     Master: 4 GB heap
>     RegionServer: 8 GB heap, 24 GB bucket cache offheap engine
>
>   1x EC2 cc2.8xlarge
>     YCSB
>
> *Configuration*
>
>   Hadoop and HBase default site configuration, except:
>    - dfs.client.read.shortcircuit=true
>    - dfs.client.read.shortcircuit.skip.checksum=true
>    - dfs.client.domain.socket.data.traffic=true
>    - dfs.domain.socket.path=...
>    - dfs.datanode.max.xcievers = 16384
>    - dfs.namenode.handler.count = 100
>    - ipc.server.tcpnodelay=true
>    - hbase.ipc.client.tcpnodelay=true
>    - hbase.bucketcache.ioengine=offheap
>    - hbase.bucketcache.size=0.9
>
> *Methodology*
>
>  Setup:
>
>    0. Start cluster
>    1. shell: create "seed", { NAME=>"u", COMPRESSION=>"snappy" }
>    2. YCSB:  Preload 100 million rows into table "seed"
>    3. shell: compact "seed"
>    4. Wait for compaction to complete
>    5. shell: create_snapshot "seed", "seed_snap"
>    6. shell: disable "seed"
>
>   For each test:
>
>    7. shell: clone_snapshot "seed_snap", "test"
>    8. YCSB:  Run test -p operationcount=10000000 -threads 32 -target 50000
> (clamp at ~10k ops/server/sec)
>    9. shell: disable "test"
>   10. shell: drop "test"
>
> *Workload A*
>
> *  0.98.0*
>
>   [OVERALL], RunTime(ms), 200313.0
>   [OVERALL], Throughput(ops/sec), 49921.87226989761
>   [READ], Operations, 5002419
>   [READ], AverageLatency(us), 1136.5890318264023
>   [READ], MinLatency(us), 281
>   [READ], MaxLatency(us), 1134224
>   [READ], 95thPercentileLatency(ms), 1
>   [READ], 99thPercentileLatency(ms), 2
>   [UPDATE], Operations, 4997613
>   [UPDATE], AverageLatency(us), 6.608387844356896
>   [UPDATE], MinLatency(us), 0
>   [UPDATE], MaxLatency(us), 724894
>   [UPDATE], 95thPercentileLatency(ms), 0
>   [UPDATE], 99thPercentileLatency(ms), 0
>
> *  0.98.1*
>
>   [OVERALL], RunTime(ms), 200323.0
>   [OVERALL], Throughput(ops/sec), 49919.38020097542
>   [READ], Operations, 5001366
>   [READ], AverageLatency(us), 1157.195526182247
>   [READ], MinLatency(us), 311
>   [READ], MaxLatency(us), 1116840
>   [READ], 95thPercentileLatency(ms), 1
>   [READ], 99thPercentileLatency(ms), 2
>   [UPDATE], Operations, 4998666
>   [UPDATE], AverageLatency(us), 6.185563908450774
>   [UPDATE], MinLatency(us), 0
>   [UPDATE], MaxLatency(us), 725483
>   [UPDATE], 95thPercentileLatency(ms), 0
>   [UPDATE], 99thPercentileLatency(ms), 0
>
> *Workload B*
>
> *  0.98.0*
>
>   [OVERALL], RunTime(ms), 264508.0
>   [OVERALL], Throughput(ops/sec), 37806.039892933295
>   [READ], Operations, 9499452
>   [READ], AverageLatency(us), 880.201064545618
>   [READ], MinLatency(us), 290
>   [READ], MaxLatency(us), 1160947
>   [READ], 95thPercentileLatency(ms), 1
>   [READ], 99thPercentileLatency(ms), 1
>   [UPDATE], Operations, 500580
>   [UPDATE], AverageLatency(us), 12.54600063925846
>   [UPDATE], MinLatency(us), 1
>   [UPDATE], MaxLatency(us), 741148
>   [UPDATE], 95thPercentileLatency(ms), 0
>   [UPDATE], 99thPercentileLatency(ms), 0
>
> *  0.98.1*
>
>   [OVERALL], RunTime(ms), 265789.0
>   [OVERALL], Throughput(ops/sec), 37623.8294286069
>   [READ], Operations, 9499661
>   [READ], AverageLatency(us), 884.9811973290415
>   [READ], MinLatency(us), 273
>   [READ], MaxLatency(us), 1186243
>   [READ], 95thPercentileLatency(ms), 1
>   [READ], 99thPercentileLatency(ms), 1
>   [UPDATE], Operations, 500371
>   [UPDATE], AverageLatency(us), 16.71922433554303
>   [UPDATE], MinLatency(us), 1
>   [UPDATE], MaxLatency(us), 923659
>   [UPDATE], 95thPercentileLatency(ms), 0
>   [UPDATE], 99thPercentileLatency(ms), 0
>
> *Workload C*
>
> *  0.98.0*
>
>   [OVERALL], RunTime(ms), 275967.0
>   [OVERALL], Throughput(ops/sec), 36236.2166490921
>   [READ], Operations, 10000000
>   [READ], AverageLatency(us), 875.3169807
>   [READ], MinLatency(us), 283
>   [READ], MaxLatency(us), 1123204
>   [READ], 95thPercentileLatency(ms), 1
>   [READ], 99thPercentileLatency(ms), 1
>
> *  0.98.1*
>
>   [OVERALL], RunTime(ms), 274182.0
>   [OVERALL], Throughput(ops/sec), 36472.124355355205
>   [READ], Operations, 10000000
>   [READ], AverageLatency(us), 868.9167299
>   [READ], MinLatency(us), 287
>   [READ], MaxLatency(us), 1082448
>   [READ], 95thPercentileLatency(ms), 1
>   [READ], 99thPercentileLatency(ms), 1
>
> *Workload D*
>
> *  0.98.0*
>
>   [OVERALL], RunTime(ms), 430518.0
>   [OVERALL], Throughput(ops/sec), 23227.83251803641
>   [INSERT], Operations, 501258
>   [INSERT], AverageLatency(us), 18.74352529036943
>   [INSERT], MinLatency(us), 5
>   [INSERT], MaxLatency(us), 729485
>   [INSERT], 95thPercentileLatency(ms), 0
>   [INSERT], 99thPercentileLatency(ms), 0
>   [READ], Operations, 9498742
>   [READ], AverageLatency(us), 1435.8229302364462
>   [READ], MinLatency(us), 278
>   [READ], MaxLatency(us), 1086256
>   [READ], 95thPercentileLatency(ms), 3
>   [READ], 99thPercentileLatency(ms), 13
>
> *  0.98.1*
>
>   [OVERALL], RunTime(ms), 473086.0
>   [OVERALL], Throughput(ops/sec), 21137.805811205573
>   [INSERT], Operations, 500280
>   [INSERT], AverageLatency(us), 18.910943871431996
>   [INSERT], MinLatency(us), 5
>   [INSERT], MaxLatency(us), 741632
>   [INSERT], 95thPercentileLatency(ms), 0
>   [INSERT], 99thPercentileLatency(ms), 0
>   [READ], Operations, 9499720
>   [READ], AverageLatency(us), 1574.2442979372024
>   [READ], MinLatency(us), 285
>   [READ], MaxLatency(us), 1142568
>   [READ], 95thPercentileLatency(ms), 6
>   [READ], 99thPercentileLatency(ms), 15
>
> *Workload E*
>
> *  0.98.0*
>
>   [OVERALL], RunTime(ms), 1741385.0
>   [OVERALL], Throughput(ops/sec), 5742.555494620661
>   [INSERT], Operations, 498580
>   [INSERT], AverageLatency(us), 12.785394520438045
>   [INSERT], MinLatency(us), 5
>   [INSERT], MaxLatency(us), 47371
>   [INSERT], 95thPercentileLatency(ms), 0
>   [INSERT], 99thPercentileLatency(ms), 0
>   [SCAN], Operations, 9501420
>   [SCAN], AverageLatency(us), 5832.021051169194
>   [SCAN], MinLatency(us), 850
>   [SCAN], MaxLatency(us), 18383187
>   [SCAN], 95thPercentileLatency(ms), 11
>   [SCAN], 99thPercentileLatency(ms), 22
>
> *  0.98.1*
>
>   [OVERALL], RunTime(ms), 1626498.0
>   [OVERALL], Throughput(ops/sec), 6148.178479161978
>   [INSERT], Operations, 498248
>   [INSERT], AverageLatency(us), 14.589842809203448
>   [INSERT], MinLatency(us), 5
>   [INSERT], MaxLatency(us), 726310
>   [INSERT], 95thPercentileLatency(ms), 0
>   [INSERT], 99thPercentileLatency(ms), 0
>   [SCAN], Operations, 9501752
>   [SCAN], AverageLatency(us), 5451.874070381968
>   [SCAN], MinLatency(us), 847
>   [SCAN], MaxLatency(us), 1278273
>   [SCAN], 95thPercentileLatency(ms), 12
>   [SCAN], 99thPercentileLatency(ms), 26
>
> *Workload F*
>
> *  0.98.0*
>
>   [OVERALL], RunTime(ms), 376552.0
>   [OVERALL], Throughput(ops/sec), 26556.75710127685
>   [READ], Operations, 10000000
>   [READ], AverageLatency(us), 1192.262719
>   [READ], MinLatency(us), 293
>   [READ], MaxLatency(us), 1108215
>   [READ], 95thPercentileLatency(ms), 1
>   [READ], 99thPercentileLatency(ms), 2
>   [READ-MODIFY-WRITE], Operations, 4999036
>   [READ-MODIFY-WRITE], AverageLatency(us), 1198.5333812359022
>   [READ-MODIFY-WRITE], MinLatency(us), 299
>   [READ-MODIFY-WRITE], MaxLatency(us), 1070294
>   [READ-MODIFY-WRITE], 95thPercentileLatency(ms), 1
>   [READ-MODIFY-WRITE], 99thPercentileLatency(ms), 2
>   [READ-MODIFY-WRITE], 0, 2018024
>   [UPDATE], Operations, 4999068
>   [UPDATE], AverageLatency(us), 3.940626532785711
>   [UPDATE], MinLatency(us), 1
>   [UPDATE], MaxLatency(us), 155147
>   [UPDATE], 95thPercentileLatency(ms), 0
>   [UPDATE], 99thPercentileLatency(ms), 0
>
> *  0.98.1*
>
>   [OVERALL], RunTime(ms), 385273.0
>   [OVERALL], Throughput(ops/sec), 25955.621079078992
>   [READ], Operations, 10000000
>   [READ], AverageLatency(us), 1218.9608977
>   [READ], MinLatency(us), 296
>   [READ], MaxLatency(us), 1128785
>   [READ], 95thPercentileLatency(ms), 1
>   [READ], 99thPercentileLatency(ms), 3
>   [READ-MODIFY-WRITE], Operations, 5001281
>   [READ-MODIFY-WRITE], AverageLatency(us), 1226.2466698031965
>   [READ-MODIFY-WRITE], MinLatency(us), 300
>   [READ-MODIFY-WRITE], MaxLatency(us), 1129077
>   [READ-MODIFY-WRITE], 95thPercentileLatency(ms), 1
>   [READ-MODIFY-WRITE], 99thPercentileLatency(ms), 3
>   [UPDATE], Operations, 5001313
>   [UPDATE], AverageLatency(us), 4.26973976633736
>   [UPDATE], MinLatency(us), 1
>   [UPDATE], MaxLatency(us), 120701
>   [UPDATE], 95thPercentileLatency(ms), 0
>   [UPDATE], 99thPercentileLatency(ms), 0
>
>
> --
> Best regards,
>
>    - Andy
>
> Problems worthy of attack prove their worth by hitting back. - Piet Hein
> (via Tom White)
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message