[
https://issues.apache.org/jira/browse/CASSANDRA-8714?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14309882#comment-14309882
]
Robert Stupp commented on CASSANDRA-8714:
-----------------------------------------
I did some off-heap allocation microbenchmarking and published a blog post
about it:
http://hardcodejavadog.blogspot.de/2015/02/comparing-java-off-heap-memory.html
It contains measurements on
* AMD CPU (6 cores, a bit old, but ok) / Linux
* Core i7 Ive Bridge (4 cores + HT), latest OS X
* m3.2xlarge (8 cores)
* c3.4xlarge (16 cores)
* c4.8xlarge (36 cores)
Some things to point out:
# jemalloc via JNA library can hit some "busy deadlock" that effectively leaves
threads hanging (and consuming some CPU - feels like a busy spin lock) - filed
[JNA bug #396|https://github.com/twall/jna/issues/396]
# jemalloc via JNA library is the slowest possible method on Linux and OS X
# jemalloc via JNA's own malloc() + free() methods with jemalloc preloaded is
the fastest possible thing
# Unsafe with jemalloc preloaded is not bad, too
# That good old AMD CPU has a quite nice allocation throughput ;)
> row-cache: use preloaded jemalloc w/ Unsafe
> -------------------------------------------
>
> Key: CASSANDRA-8714
> URL: https://issues.apache.org/jira/browse/CASSANDRA-8714
> Project: Cassandra
> Issue Type: Sub-task
> Reporter: Robert Stupp
> Assignee: Robert Stupp
> Fix For: 3.0
>
> Attachments: 8714.txt
>
>
> Using jemalloc via Java's {{Unsafe}} is a better alternative on Linux than
> using jemalloc via JNA.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
