[ 
https://issues.apache.org/jira/browse/MAPREDUCE-6923?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16120767#comment-16120767
 ] 

Ravi Prakash commented on MAPREDUCE-6923:
-----------------------------------------

Hi Robert!

Here's my reasoning about this patch. Sorry about being this verbose. I just 
have umm.... let's say history with the shuffle code ;-) : 
# When {{shuffleBufferSize <= trans}}, then behavior is exactly the same as old 
code.
# When {{trans < shuffleBufferSize}} then 
#* if {{readSize == trans}} (i.e. the {{fileChannel.read()}} returned as many 
bytes as I wanted to transfer, {{trans}} is decremented correctly, {{position}} 
is increased correctly and the {{byteBuffer}} is flipped as usual. 
{{byteBuffer}}'s contents are written to {{target}} as usual, {{byteBuffer}} is 
cleared and then hopefully GCed never to be seen again.
#* if {{readSize < trans}} (almost the same thing as above happens, but in a 
while loop). The only change this patch makes is that the {{byteBuffer}} may be 
smaller than before this patch, but it doesn't matter because its big enough 
for the number of bytes we need to transfer.
#* if {{readSize > trans}} This shouldn't happen any more since 
{{byteBuffer}}'s size is {{trans}} . However this is still not dead code 
because we need it for the first case (when {{shuffleBufferSize <= trans}})

As much as I would have liked another review to calm myself, I am fairly 
confident this is fine. Please let me know if the reasoning above is incorrect 
in any manner.

Committing shortly

> YARN Shuffle I/O for small partitions
> -------------------------------------
>
>                 Key: MAPREDUCE-6923
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-6923
>             Project: Hadoop Map/Reduce
>          Issue Type: Improvement
>         Environment: Observed in Hadoop 2.7.3 and above (judging from the 
> source code of future versions), and Ubuntu 16.04.
>            Reporter: Robert Schmidtke
>            Assignee: Robert Schmidtke
>         Attachments: MAPREDUCE-6923.00.patch, MAPREDUCE-6923.01.patch
>
>
> When a job configuration results in small partitions read by each reducer 
> from each mapper (e.g. 65 kilobytes as in my setup: a 
> [TeraSort|https://github.com/apache/hadoop/blob/branch-2.7.3/hadoop-mapreduce-project/hadoop-mapreduce-examples/src/main/java/org/apache/hadoop/examples/terasort/TeraSort.java]
>  of 256 gigabytes using 2048 mappers and reducers each), and setting
> {code:xml}
> <property>
>   <name>mapreduce.shuffle.transferTo.allowed</name>
>   <value>false</value>
> </property>
> {code}
> then the default setting of
> {code:xml}
> <property>
>   <name>mapreduce.shuffle.transfer.buffer.size</name>
>   <value>131072</value>
> </property>
> {code}
> results in almost 100% overhead in reads during shuffle in YARN, because for 
> each 65K needed, 128K are read.
> I propose a fix in 
> [FadvisedFileRegion.java|https://github.com/apache/hadoop/blob/trunk/hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-shuffle/src/main/java/org/apache/hadoop/mapred/FadvisedFileRegion.java#L114]
>  as follows:
> {code:java}
> ByteBuffer byteBuffer = ByteBuffer.allocate(Math.min(this.shuffleBufferSize, 
> trans > Integer.MAX_VALUE ? Integer.MAX_VALUE : (int) trans));
> {code}
> e.g. 
> [here|https://github.com/apache/hadoop/compare/branch-2.7.3...robert-schmidtke:adaptive-shuffle-buffer].
>  This sets the shuffle buffer size to the minimum value of the shuffle buffer 
> size specified in the configuration (128K by default), and the actual 
> partition size (65K on average in my setup). In my benchmarks this reduced 
> the read overhead in YARN from about 100% (255 additional gigabytes as 
> described above) down to about 18% (an additional 45 gigabytes). The runtime 
> of the job remained the same in my setup.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: mapreduce-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: mapreduce-issues-h...@hadoop.apache.org

Reply via email to