[
https://issues.apache.org/jira/browse/HDFS-918?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12829368#action_12829368
]
ryan rawson commented on HDFS-918:
----------------------------------
the problem was we were using a stateful interface previously because it was
faster in scan tests, so we serialized reads within 1 RS to any given HFile.
With multiple client handler threads asking for different parts of a large
file, we get a serialized behaviour which hurts random get performance.
So we are moving back to pread, which means we will get more parallelism -
depending your table read pattern of course. But I want to get even more
parallelism, by preading multiple hfiles during a scan/get for example. This
will just up the thread pressure on the datanode.
> Use single Selector and small thread pool to replace many instances of
> BlockSender for reads
> --------------------------------------------------------------------------------------------
>
> Key: HDFS-918
> URL: https://issues.apache.org/jira/browse/HDFS-918
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: data-node
> Reporter: Jay Booth
> Fix For: 0.22.0
>
> Attachments: hdfs-918-20100201.patch, hdfs-918-20100203.patch,
> hdfs-multiplex.patch
>
>
> Currently, on read requests, the DataXCeiver server allocates a new thread
> per request, which must allocate its own buffers and leads to
> higher-than-optimal CPU and memory usage by the sending threads. If we had a
> single selector and a small threadpool to multiplex request packets, we could
> theoretically achieve higher performance while taking up fewer resources and
> leaving more CPU on datanodes available for mapred, hbase or whatever. This
> can be done without changing any wire protocols.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
