[
https://issues.apache.org/jira/browse/HTRACE-298?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15014813#comment-15014813
]
Colin Patrick McCabe commented on HTRACE-298:
---------------------------------------------
bq. The patch contains a lot of fixes for tests, metrics and variable names.
Because datastore serialization is not trivial change, should we split the
patch and focus only on the serialization here?
The test fixes are really trivial, basically replacing the channel with a
semaphore. I will post a patch on HTRACE-301 to do that stuff.
The changes to span ingestion and metrics are really somewhat tied together so
I'd rather not split them unless it's absolutely necessary. Basically, it is a
lot more efficient to increment metrics in batches, and that is only possible
when spans are ingested in batches.
bq. Does that mean buffer of channel was not enough and adding the Semaphore
free us from worrying about the buffer size?
Buffering the channel is usually enough, but some of the performance tests
create hundreds of thousands or millions of spans. I didn't want to create
such huge channel buffers.
bq. Posts takes int64 as argument but Waits takes int.
Yeah, let's make them both {{int64}}.
bq. It looks inconsistent that Post checks sem.count > 0 but Posts does not.
Yeah, let's add this small optimization to {{Posts}}.
> htraced: improve datastore serialization
> ----------------------------------------
>
> Key: HTRACE-298
> URL: https://issues.apache.org/jira/browse/HTRACE-298
> Project: HTrace
> Issue Type: Improvement
> Components: htraced
> Affects Versions: 4.0
> Reporter: Colin Patrick McCabe
> Assignee: Colin Patrick McCabe
> Attachments: HTRACE-298.001.patch, HTRACE-298.002.patch
>
>
> htraced: speed up datastore serialization
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
