[
https://issues.apache.org/jira/browse/HADOOP-11226?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14326215#comment-14326215
]
Chris Nauroth commented on HADOOP-11226:
----------------------------------------
bq. I also think a lot of people would disagree that latency is more important
than throughput for HDFS RPC.
I think it's notable that the change is only on RPC and not on data transfer
protocol, where I'd expect most people would favor throughput.
bq. It seems like we should make this configurable (and not on by default)
since we don't know exactly what it will do on most networks.
I generally think we have too many configuration flags in Hadoop already, but
on this one, I have to agree. If it works out well in real-world practice,
then we could make it the default at some point later, perhaps on a major
version boundary.
> ipc.Client has to use setTrafficClass() with IPTOS_LOWDELAY|IPTOS_RELIABILITY
> -----------------------------------------------------------------------------
>
> Key: HADOOP-11226
> URL: https://issues.apache.org/jira/browse/HADOOP-11226
> Project: Hadoop Common
> Issue Type: Bug
> Components: ipc
> Affects Versions: 2.6.0
> Reporter: Gopal V
> Assignee: Gopal V
> Labels: Infiniband
> Attachments: HADOOP-11226.1.patch, HADOOP-11226.2.patch
>
>
> During heavy shuffle, packet loss for IPC packets was observed from a machine.
> Avoid packet-loss and speed up transfer by using 0x14 QOS bits for the
> packets.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
