[
https://issues.apache.org/jira/browse/HADOOP-11226?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14290067#comment-14290067
]
Gopal V commented on HADOOP-11226:
----------------------------------
Thanks [~cnauroth]. I think this needs testing beyond what I've done - my
test-setup is a single rack with has 10G ethernet - using both H-TCP & Cubic
congestion controls.
This definitely needs more testing, should be even better when encountering
cross-rack packet loss.
In particular, I'd love to see this patch tested with JDK7 Infiniband native
support.
Hadoop IPC/DataStreamer/Shuffle can already use Infiniband zero-copy mode by
adding JVM opt args configuration options {{-Dcom.sun.sdp.conf=/etc/sdp.conf
-Dcom.sun.sdp.debug=/tmp/sdp-debug.log}}, provided the OS/Hardware has support.
> 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)
