[
https://issues.apache.org/jira/browse/HBASE-15213?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15133842#comment-15133842
]
stack commented on HBASE-15213:
-------------------------------
Ok. I got below numbers. This patch is great because it does away w/ the need
of hbase.increment.fast.but.narrow.consistency on branch 1.0 and 1.1 (and
branch-1.2 -- and it seems it is not needed there anyways). It also is better
because it addresses appends and checkAndPut (I'm pretty sure -- I will
verify). Let me apply this patch in the morning. I want to update release
notes to sing praises of this fix and to do proper messaging; it'll easy for
users to get confused. Let me try and take care of that.
Nice one [~junegunn] You sure there ain't other places you'd like to go digging
in (smile).
Before HBASE-8763: d6cc2fb
{code}
2016-02-05 00:18:58,665 WARN [pool-1-thread-78] hbase.HBaseConfiguration:
hbase.regionserver.global.memstore.upperLimit is deprecated by
hbase.regionserver.global.memstore.size
2016-02-05 00:19:55,814 INFO [main] hbase.IncrementPerformanceTest:
75th=5.329, 95th=9.355449999999998, 99th=19.42884
{code}
Tip of 1.0 branch
{code}
2016-02-05 00:22:44,475 INFO [pool-1-thread-78-SendThread(localhost:2181)]
zookeeper.ClientCnxn: Session establishment complete on server
localhost/0:0:0:0:0:0:0:1:2181, sessionid = 0x152b085e9ae0007, negotiated
timeout = 40000
2016-02-05 00:26:02,829 INFO [main] hbase.IncrementPerformanceTest:
75th=26.22475, 95th=36.996399999999994, 99th=44.89627000000002
{code}
Tip of 1.0 branch with hbase.increment.fast.but.narrow.consistency
{code}
2016-02-05 00:28:57,215 INFO [pool-1-thread-4-SendThread(localhost:2181)]
zookeeper.ClientCnxn: Session establishment complete on server
localhost/127.0.0.1:2181, sessionid = 0x152b08c36410006, negotiated timeout =
40000
2016-02-05 00:29:50,879 INFO [main] hbase.IncrementPerformanceTest:
75th=5.1594999999999995, 95th=11.920049999999998, 99th=34.12252000000004
{code}
Tip of 1.0 with patch applied
{code}
2016-02-05 00:35:27,268 INFO [pool-1-thread-64-SendThread(localhost:2181)]
zookeeper.ClientCnxn: Session establishment complete on server
localhost/0:0:0:0:0:0:0:1:2181, sessionid = 0x152b091faec0006, negotiated
timeout = 40000
2016-02-05 00:36:20,006 INFO [main] hbase.IncrementPerformanceTest:
75th=5.039, 95th=9.919499999999992, 99th=20.12136000000002
{code}
> Fix increment performance regression caused by HBASE-8763 on branch-1.0
> -----------------------------------------------------------------------
>
> Key: HBASE-15213
> URL: https://issues.apache.org/jira/browse/HBASE-15213
> Project: HBase
> Issue Type: Bug
> Components: Performance
> Reporter: Junegunn Choi
> Assignee: Junegunn Choi
> Attachments: HBASE-15213-increment.png, HBASE-15213.branch-1.0.patch
>
>
> This is an attempt to fix the increment performance regression caused by
> HBASE-8763 on branch-1.0.
> I'm aware that hbase.increment.fast.but.narrow.consistency was added to
> branch-1.0 (HBASE-15031) to address the issue and a separate work is ongoing
> on master branch, but anyway, this is my take on the problem.
> I read through HBASE-14460 and HBASE-8763 but it wasn't clear to me what
> caused the slowdown but I could indeed reproduce the performance regression.
> Test setup:
> - Server: 4-core Xeon 2.4GHz Linux server running mini cluster (100 handlers,
> JDK 1.7)
> - Client: Another box of the same spec
> - Increments on random 10k records on a single-region table, recreated every
> time
> Increment throughput (TPS):
> || Num threads || Before HBASE-8763 (d6cc2fb) || branch-1.0 || branch-1.0
> (narrow-consistency) ||
> || 1 | 2661 | 2486 | 2359 |
> || 2 | 5048 | 5064 | 4867 |
> || 4 | 7503 | 8071 | 8690 |
> || 8 | 10471 | 10886 | 13980 |
> || 16 | 15515 | 9418 | 18601 |
> || 32 | 17699 | 5421 | 20540 |
> || 64 | 20601 | 4038 | 25591 |
> || 96 | 19177 | 3891 | 26017 |
> We can clearly observe that the throughtput degrades as we increase the
> number of concurrent requests, which led me to believe that there's severe
> context switching overhead and I could indirectly confirm that suspicion with
> cs entry in vmstat output. branch-1.0 shows a much higher number of context
> switches even with much lower throughput.
> Here are the observations:
> - WriteEntry in the writeQueue can only be removed by the very handler that
> put it, only when it is at the front of the queue and marked complete.
> - Since a WriteEntry is marked complete after the wait-loop, only one entry
> can be removed at a time.
> - This stringent condition causes O(N^2) context switches where n is the
> number of concurrent handlers processing requests.
> So what I tried here is to mark WriteEntry complete before we go into
> wait-loop. With the change, multiple WriteEntries can be shifted at a time
> without context switches. I changed writeQueue to LinkedHashSet since fast
> containment check is needed as WriteEntry can be removed by any handler.
> The numbers look good, it's virtually identical to pre-HBASE-8763 era.
> || Num threads || branch-1.0 with fix ||
> || 1 | 2459 |
> || 2 | 4976 |
> || 4 | 8033 |
> || 8 | 12292 |
> || 16 | 15234 |
> || 32 | 16601 |
> || 64 | 19994 |
> || 96 | 20052 |
> So what do you think about it? Please let me know if I'm missing anything.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
