hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Doug Cutting (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-3556) Substitute the synchronized code in MD5Hash to avoid lock contention. Use ThreadLocal instead.
Date Mon, 16 Jun 2008 16:51:46 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-3556?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12605344#action_12605344
] 

Doug Cutting commented on HADOOP-3556:
--------------------------------------

You found a bug in Hadoop's workflow.  The submitter can push an issue into the "In Progress"
state, but only the assignee can push it from that state into "patch available".  I assigned
the issue to you, so now you should be able to do this.

(Jira workflows are very heavyweight and I don't think it is worth trying to fix this.)


> Substitute the synchronized code in MD5Hash to avoid lock contention. Use ThreadLocal
instead.
> ----------------------------------------------------------------------------------------------
>
>                 Key: HADOOP-3556
>                 URL: https://issues.apache.org/jira/browse/HADOOP-3556
>             Project: Hadoop Core
>          Issue Type: Improvement
>            Reporter: Iván de Prado
>            Assignee: Iván de Prado
>         Attachments: 3556.patch, MD5Hash-ThreadLocal.patch
>
>
> Currently the MD5Hash class uses a singleton instance of the MessageDigester. The access
to this instance is synchronized, so MD5Hash has performance problems when used from several
Threads. I propose to substitute the singleton instance by a TheadLocal instances cache. I
will provide a patch.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message