[
https://issues.apache.org/jira/browse/HDFS-7847?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15015171#comment-15015171
]
Colin Patrick McCabe commented on HDFS-7847:
--------------------------------------------
bq. what are you measuring
NNThroughput, measured in ops per second.
bq. how you set up the nodes
Any way you like, depending on what you are benchmarking. There is no special
setup needed.
bq. are there any particular RPC improvements or alternative implementations
you are comparing
Somewhat recently, I did some benchmarks to figure out how to tune the NN to
get the most ops per second. Actually on my particular hardware, fsync turned
out to be a bottleneck, since hard disk write caching was turned off. In
general, fsync is not a bottleneck, but it was in my specific case. More
generally, we should pay attention to benchmarks like these to make sure our
performance doesn't drop.
> Modify NNThroughputBenchmark to be able to operate on a remote NameNode
> -----------------------------------------------------------------------
>
> Key: HDFS-7847
> URL: https://issues.apache.org/jira/browse/HDFS-7847
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 2.8.0
> Reporter: Colin Patrick McCabe
> Assignee: Charles Lamb
> Fix For: 2.8.0
>
> Attachments: HDFS-7847.000.patch, HDFS-7847.001.patch,
> HDFS-7847.002.patch, HDFS-7847.003.patch, HDFS-7847.004.patch,
> HDFS-7847.005.patch, make_blocks.tar.gz
>
>
> Modify NNThroughputBenchmark to be able to operate on a NN that is not in
> process. A followon Jira will modify it some more to allow quantifying native
> and java heap sizes, and some latency numbers.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
