[
https://issues.apache.org/jira/browse/PHOENIX-1711?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14353185#comment-14353185
]
Gabriel Reid commented on PHOENIX-1711:
---------------------------------------
{quote}
I'm curious about one thing, though - do you think there's any overhead in
calling CallRunner.run() per row versus once per batch?
{quote}
I would assume that it's lightweight enough that it won't make a measurable
difference, but that's just my guess.
Actually, I was curious about why that the call to CallRunner.run was needed at
all in the map method of CSVUpsertExecutor -- at least when it's being used
within MapReduce, it should be fine to just set the context classloader once
(which is what CallRunner does, from what I see). From what I recall, the
reason for having the CallRunner system is to "fix" the context classloader
when running inside of JDBC tooling that dynamically loads Phoenix, but I
wouldn't think that there would be ever be a problem when running via the bulk
loader.
> Improve performance of CSV loader
> ---------------------------------
>
> Key: PHOENIX-1711
> URL: https://issues.apache.org/jira/browse/PHOENIX-1711
> Project: Phoenix
> Issue Type: Bug
> Reporter: James Taylor
> Attachments: PHOENIX-1711.patch
>
>
> Here is a break-up of percentage execution time for some of the steps inthe
> mapper:
> csvParser: 18%
> csvUpsertExecutor.execute(ImmutableList.of(csvRecord)): 39%
> PhoenixRuntime.getUncommittedDataIterator(conn, true): 9%
> while (uncommittedDataIterator.hasNext()): 15%
> Read IO & custom processing: 19%
> See details here: http://s.apache.org/6rl
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
