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 "Amar Kamat (JIRA)" <ji...@apache.org> on 2008/05/07 08:21:55 UTC

[jira] Issue Comment Edited: (HADOOP-3116) Investigate whether the array in the JobInProgress that holds TIP references can be removed

    [ https://issues.apache.org/jira/browse/HADOOP-3116?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12594790#action_12594790 ] 

amar_kamat edited comment on HADOOP-3116 at 5/6/08 11:21 PM:
-------------------------------------------------------------

One simple thing we can do is to use a {{MAP}} from _id-within-job_ to _TIP_ instead of an array.

      was (Author: amar_kamat):
    One simple thing we can do is to use a {{MAP}} from _attempt-id_ to _TIP_ instead of an array.
  
> Investigate whether the array in the JobInProgress that holds TIP references can be removed
> -------------------------------------------------------------------------------------------
>
>                 Key: HADOOP-3116
>                 URL: https://issues.apache.org/jira/browse/HADOOP-3116
>             Project: Hadoop Core
>          Issue Type: Improvement
>          Components: mapred
>    Affects Versions: 0.17.0
>            Reporter: Devaraj Das
>             Fix For: 0.18.0
>
>
> The array, in JobInProgress, that holds the references to TIPs is required now only to serve the clients (web UI via JSPs and JobClient APIs, that traverse the array and create the output). The array can now be removed since HADOOP-2119 introduces some datastructures for running/non-running TIPs and we could probably add some more if required (e.g., for completed TIPs). That way, we will gain when we have large jobs (many tasks) in the sense that we don't have to go through the big array every time a client request is made. We could instead iterate over the datastructures. Also, we should do these traversals without locking the JobTracker to avoid cases where long traversal leads to loss of heartbeats, etc. Some staleness at the client side (in favor of improving the JobTracker's performance) is probably okay.

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