[
https://issues.apache.org/jira/browse/HDFS-14162?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16767374#comment-16767374
]
Erik Krogen edited comment on HDFS-14162 at 2/13/19 4:38 PM:
-------------------------------------------------------------
I ran a JMH benchmark to see about performance differences (
[^ReflectionBenchmark.java] ). It looks like the case where the method exists
is much faster whether using direct invocation or using a method lookup, and
the direct invocation is faster whether or not the method exists:
{code:java}
Benchmark Mode Cnt
Score Error Units
ReflectionBenchmark.testDirectInvocationWhenExists thrpt 25
90940155.095 ± 566784.364 ops/s
ReflectionBenchmark.testDirectInvocationWhenNotExists thrpt 25
397612.090 ± 7832.119 ops/s
ReflectionBenchmark.testGetMethodWhenExists thrpt 25
5377201.019 ± 174039.222 ops/s
ReflectionBenchmark.testGetMethodWhenNotExists thrpt 25
329229.282 ± 6865.177 ops/s
{code}
Attached v003 patch with license update. Also checked
{{TestBalancerWithMultipleNameNodes}} locally, and it passes fine. I've seen
this test be flaky (timeout) in precommit builds recently.
was (Author: xkrogen):
I ran a JMH benchmark to see about performance differences (
[^ReflectionBenchmark.java] ). It looks like the case where the method exists
is much faster whether using direct invocation or using a method lookup, and
the direct invocation is faster whether or not the method exists:
{code:java}
Benchmark Mode Cnt
Score Error Units
ReflectionBenchmark.testDirectInvocationWhenExists thrpt 25
90940155.095 ± 566784.364 ops/s
ReflectionBenchmark.testDirectInvocationWhenNotExists thrpt 25
397612.090 ± 7832.119 ops/s
ReflectionBenchmark.testGetMethodWhenExists thrpt 25
5377201.019 ± 174039.222 ops/s
ReflectionBenchmark.testGetMethodWhenNotExists thrpt 25
329229.282 ± 6865.177 ops/s
{code}
> Balancer should work with ObserverNode
> --------------------------------------
>
> Key: HDFS-14162
> URL: https://issues.apache.org/jira/browse/HDFS-14162
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Reporter: Konstantin Shvachko
> Assignee: Erik Krogen
> Priority: Major
> Attachments: HDFS-14162-HDFS-12943.wip0.patch, HDFS-14162.000.patch,
> HDFS-14162.001.patch, HDFS-14162.002.patch, HDFS-14162.003.patch,
> ReflectionBenchmark.java, testBalancerWithObserver-3.patch,
> testBalancerWithObserver.patch
>
>
> Balancer provides a substantial RPC load on NameNode. It would be good to
> divert Balancer RPCs {{getBlocks()}}, etc. to ObserverNode. The main problem
> is that Balancer uses {{NamenodeProtocol}}, while ORPP currently supports
> only {{ClientProtocol}}.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
