[
https://issues.apache.org/jira/browse/HADOOP-1965?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12538472
]
Amar Kamat commented on HADOOP-1965:
------------------------------------
Results contd.. The setup is as follows
----
System config :
* # Cores : 4
* Heap size : 2GB
----
Hadoop config
* Data source: Random text using random-text-writer
* Key size : single word
* Value size : empty
* Data size : ~392mb(uncompressed) ~75mb(compressed)
* Input type : Gzip
* DFS-block-size : 128mb
* # nodes : 1(tasktracker + datanode) and 1(NN + JT)
* Job type : wordcount
* io.sort.mb : 180
* # maps : 2 maps concurrently
----
Results : We observed 16.79% improvement with threaded spill.
----
comments ?
> Handle map output buffers better
> --------------------------------
>
> Key: HADOOP-1965
> URL: https://issues.apache.org/jira/browse/HADOOP-1965
> Project: Hadoop
> Issue Type: Improvement
> Components: mapred
> Reporter: Devaraj Das
> Assignee: Amar Kamat
> Attachments: 1965_single_proc_150mb_gziped.jpeg,
> 1965_single_proc_150mb_gziped.pdf, 1965_single_proc_150mb_gziped_breakup.png
>
>
> Today, the map task stops calling the map method while sort/spill is using
> the (single instance of) map output buffer. One improvement that can be done
> to improve performance of the map task is to have another buffer for writing
> the map outputs to, while sort/spill is using the first buffer.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
