You are viewing a plain text version of this content. The canonical link for it is here.

Posted to common-dev@hadoop.apache.org by "Raghu Angadi (JIRA)" <ji...@apache.org> on 2008/10/11 02:31:44 UTC

[jira] Commented: (HADOOP-4394) DFSClient cpu overhead is too high

    [ https://issues.apache.org/jira/browse/HADOOP-4394?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12638693#action_12638693 ] 

Raghu Angadi commented on HADOOP-4394:
--------------------------------------

HADOOP-3205 would help reduce a data copy on the client.. which could reduce 25-30% cpu on DFSClient. Using Adler32 for checksums a Bryan Duxbury suggested could further reduce CPU on DFSClient.

> DFSClient cpu overhead is too high
> ----------------------------------
>
>                 Key: HADOOP-4394
>                 URL: https://issues.apache.org/jira/browse/HADOOP-4394
>             Project: Hadoop Core
>          Issue Type: Improvement
>          Components: dfs
>            Reporter: Runping Qi
>
> When we do dfs throughput test using hadoop dfs -cat, we have observed that the client side cpu usage is very high, 3 to five times that of a data node serving the file.
> Before 0.18, the data node cpu usage was equally high, and this problem is fixed since 0.18. However, the client side problem still exists.

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