[
https://issues.apache.org/jira/browse/HADOOP-3164?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Raghu Angadi updated HADOOP-3164:
---------------------------------
Attachment: HADOOP-3164.patch
Updated patch includes feedback from Konstantin. Also improves documentation.
Regd Sam's experiment on Mac OS :
8x is very surprising. Sam sent me iostat during the test and the main
difference seems to be that size of each read from the disk with out the patch
is around 350 KB and just 4 KB with the patch. The latest patch uses real
socket channel with transferTo() on all the platforms. Please check if it
improves the test.
> Use FileChannel.transferTo() when data is read from DataNode.
> -------------------------------------------------------------
>
> Key: HADOOP-3164
> URL: https://issues.apache.org/jira/browse/HADOOP-3164
> Project: Hadoop Core
> Issue Type: Improvement
> Components: dfs
> Reporter: Raghu Angadi
> Assignee: Raghu Angadi
> Fix For: 0.18.0
>
> Attachments: HADOOP-3164.patch, HADOOP-3164.patch, HADOOP-3164.patch,
> HADOOP-3164.patch, HADOOP-3164.patch
>
>
> HADOOP-2312 talks about using FileChannel's
> [{{transferTo()}}|http://java.sun.com/javase/6/docs/api/java/nio/channels/FileChannel.html#transferTo(long,%20long,%20java.nio.channels.WritableByteChannel)]
> and
> [{{transferFrom()}}|http://java.sun.com/javase/6/docs/api/java/nio/channels/FileChannel.html#transferFrom(java.nio.channels.ReadableByteChannel,%20long,%20long)]
> in DataNode.
> At the time DataNode neither used NIO sockets nor wrote large chunks of
> contiguous block data to socket. Hadoop 0.17 does both when data is seved to
> clients (and other datanodes). I am planning to try using transferTo() in the
> trunk. This might reduce DataNode's cpu by another 50% or more.
> Once HADOOP-1702 is committed, we can look into using transferFrom().
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
