[
https://issues.apache.org/jira/browse/STORM-1632?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15212231#comment-15212231
]
ASF GitHub Bot commented on STORM-1632:
---------------------------------------
Github user arunmahadevan commented on the pull request:
https://github.com/apache/storm/pull/1217#issuecomment-201412463
With the BasicTopology.java uploaded in the Jira I was not getting
consistent results and in some cases the throughput without event logger turned
out to be less. After going through the spout it appears that with every emit a
new Long is created.
`collector.emit(new Values(new Long(rand.nextInt(Integer.MAX_VALUE))),
messageCount);`
With the spout emitting 400+ Million tuples in 10 mins (the new Long alone
eating up 3.2 GB memory), I thought the GC could have distorted the results. So
I changed the spout to emit a fixed long value and measured the throughput and
heres what I observed.
[No
eventlogger](https://cloud.githubusercontent.com/assets/6792890/14049885/1ccdc2ae-f2de-11e5-9f50-231990d629bb.png)
753394 tuples/sec
[With
eventlogger](https://cloud.githubusercontent.com/assets/6792890/14049918/4dd0a902-f2de-11e5-8a77-5c6ba3f20520.png)
736517 tuples/sec (2.29 % hit)
[Eventlogger with the
patch](https://cloud.githubusercontent.com/assets/6792890/14049934/61fbdce4-f2de-11e5-8454-e27db755fdc3.png)
749492 tuples/sec (0.52 % hit)
The results with the [patch]
(https://github.com/arunmahadevan/storm/commit/7eae5ec9f63cee82c49980a3bedf5f0dfe4e3a8d)
to reduce the PersistentMap lookup and the modified
[BasicTopology](https://github.com/apache/storm/files/189869/BasicTopology.txt)
can be validated and if that overhead is not acceptable, we can disable the
"debug" button by default and show a tooltip on how to enable it.
> Disable event logging by default
> --------------------------------
>
> Key: STORM-1632
> URL: https://issues.apache.org/jira/browse/STORM-1632
> Project: Apache Storm
> Issue Type: Bug
> Components: storm-core
> Reporter: Roshan Naik
> Assignee: Roshan Naik
> Priority: Blocker
> Attachments: BasicTopology.java
>
>
> EventLogging has performance penalty. For a simple speed of light topology
> with a single instances of a spout and a bolt, disabling event logging
> delivers a 7% to 9% perf improvement (with acker count =1)
> Event logging can be enabled when there is need to do debug, but turned off
> by default.
> **Update:** with acker=0 the observed impact was much higher... **25%**
> faster when event loggers = 0
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
