[
https://issues.apache.org/jira/browse/CASSANDRA-8457?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14249237#comment-14249237
]
Ariel Weisberg commented on CASSANDRA-8457:
-------------------------------------------
For me the TLDR is that they look the same. I would have to run a bunch more
times restart the DB each time and then throw away the best and the worst and
then average to see if there is a difference and if there is it is small. I
want to understand why the change had 0 impact as opposed to an insignificant
one.
The number is the run, so matching numbers ran at the same time.
Trunk
||Client|Throughput|99.9%tile||
|A #1| 87497|4.5|
|B #1| 87598|4.4|
|A#2| 89089|3.7|
|B#2| 88461|4.1|
|A#3| 89002|4.4|
|B#3| 88529|4.4|
|restart|||
|A#4| 87140|3.8|
|B#4| 87434|4.2|
With changes
||Client|Throughput|99.9%tile||
|A#1| 87476|4.0|
|B#1| 87347|3.9|
|restart|||
|A#2| 87688|4.0|
|B#2| 87849|4.4|
|A#3| 87905|4.0|
|B#3| 88834|3.9|
|A#4| 88132|4.3|
|B#4| 88187|3.8|
> nio MessagingService
> --------------------
>
> Key: CASSANDRA-8457
> URL: https://issues.apache.org/jira/browse/CASSANDRA-8457
> Project: Cassandra
> Issue Type: New Feature
> Components: Core
> Reporter: Jonathan Ellis
> Assignee: Ariel Weisberg
> Labels: performance
> Fix For: 3.0
>
>
> Thread-per-peer (actually two each incoming and outbound) is a big
> contributor to context switching, especially for larger clusters. Let's look
> at switching to nio, possibly via Netty.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
