[ https://issues.apache.org/jira/browse/HADOOP-4063?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chris Douglas updated HADOOP-4063:
----------------------------------
Resolution: Won't Fix
Status: Resolved (was: Patch Available)
In explaining how to configure MapTask to a few users, many asked why the spill thresholds
weren't separately configurable. It's easy enough to accommodate the request, but there aren't
many good use cases for it.
Auto-tuning these parameters is awkward, but we'll probably get there eventually.
> Separate spill thresholds for serialization/accounting in MapTask
> -----------------------------------------------------------------
>
> Key: HADOOP-4063
> URL: https://issues.apache.org/jira/browse/HADOOP-4063
> Project: Hadoop Core
> Issue Type: Improvement
> Components: mapred
> Reporter: Chris Douglas
> Assignee: Chris Douglas
> Priority: Minor
> Fix For: 0.19.0
>
> Attachments: 4063-0.patch, 4063-1.patch
>
>
> In MapTask, there is a single parameter controlling the threshold for starting a spill
thread concurrently with collection. However, some users may want to set different thresholds
for the serialization buffer (holding record bytes) and the accounting buffer (holding record
metadata).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
|