[
https://issues.apache.org/jira/browse/HTRACE-312?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15034496#comment-15034496
]
Colin Patrick McCabe edited comment on HTRACE-312 at 12/1/15 9:34 PM:
----------------------------------------------------------------------
Using {{-cpu 1}} I get 26341 ns/op, using {{-cpu 16}} I get 24671 ns/op, using
{{-benchtime=15s}}, for {{BenchmarkDatastoreWrites}}. In general, I would
expect gains since we spend a lot of time on serialization and deserialization,
which consume CPU.
edit: this was was on a 4-core machine with hyperthreading. I guess {{-cpu 8}}
would be more appropriate here. I've seen bigger gains on some of the cluster
nodes, which have 8 hyperthreaded CPUs.
was (Author: cmccabe):
Using {{-cpu 1}} I get 26341 ns/op, using {{-cpu 16}} I get 24671 ns/op, using
{{-benchtime=15s}}, for {{BenchmarkDatastoreWrites}}. In general, I would
expect gains since we spend a lot of time on serialization and deserialization,
which consume CPU.
> htraced: if GOMAXPROCS is left at 1, set it to the number of CPUs
> -----------------------------------------------------------------
>
> Key: HTRACE-312
> URL: https://issues.apache.org/jira/browse/HTRACE-312
> Project: HTrace
> Issue Type: Improvement
> Components: htraced
> Affects Versions: 4.0
> Reporter: Colin Patrick McCabe
> Assignee: Colin Patrick McCabe
> Attachments: HTRACE-312.001.patch
>
>
> htraced: if GOMAXPROCS is left at 1, set it to the number of CPUs.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
