[
https://issues.apache.org/jira/browse/HADOOP-3943?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Hairong Kuang updated HADOOP-3943:
----------------------------------
Attachment: sortDistance.patch
It seems to me that simply removing "synchronized" should work. There is no
need to grab the read lock.
> NetworkTopology.pseudoSortByDistance does not need to be a synchronized method
> ------------------------------------------------------------------------------
>
> Key: HADOOP-3943
> URL: https://issues.apache.org/jira/browse/HADOOP-3943
> Project: Hadoop Core
> Issue Type: Improvement
> Components: dfs
> Reporter: Hairong Kuang
> Assignee: Hairong Kuang
> Fix For: 0.19.0
>
> Attachments: sortDistance.patch
>
>
> When all requests to namenode are read, most handlers block on
> NetworkTopology object lock when sorting replicas. Removing this necessary
> lock would reduce the contention in namenode and therefore improve its
> throughput.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
