[
https://issues.apache.org/jira/browse/HDFS-10326?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15256716#comment-15256716
]
Mingliang Liu commented on HDFS-10326:
--------------------------------------
Thanks for reporting this. FWIW, [HDFS-8829]and [HDFS-9259] made the TCP auto
tuning for both read and write scenario possible with non-positive config
values. I'm not sure giving up the configurability is a good idea, but
according to [HDFS-9259], the case {{c}} for cross-DC write was the fastest.
> Disable setting tcp socket send/receive buffers for write pipelines
> -------------------------------------------------------------------
>
> Key: HDFS-10326
> URL: https://issues.apache.org/jira/browse/HDFS-10326
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, hdfs
> Affects Versions: 2.6.0
> Reporter: Daryn Sharp
> Assignee: Daryn Sharp
>
> The DataStreamer and the Datanode use a hardcoded
> DEFAULT_DATA_SOCKET_SIZE=128K for the send and receive buffers of a write
> pipeline. Explicitly setting tcp buffer sizes disables tcp stack
> auto-tuning.
> The hardcoded value will saturate a 1Gb with 1ms RTT. 105Mbs at 10ms.
> Paltry 11Mbs over a 100ms long haul. 10Gb networks are underutilized.
> There should either be a configuration to completely disable setting the
> buffers, or the the setReceiveBuffer and setSendBuffer should be removed
> entirely.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
