[
https://issues.apache.org/jira/browse/ARTEMIS-2239?focusedWorklogId=190378&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-190378
]
ASF GitHub Bot logged work on ARTEMIS-2239:
-------------------------------------------
Author: ASF GitHub Bot
Created on: 26/Jan/19 10:13
Start Date: 26/Jan/19 10:13
Worklog Time Spent: 10m
Work Description: franz1981 commented on issue #2522: ARTEMIS-2239
Zero-copy NIO/MAPPED TimedBuffer
URL: https://github.com/apache/activemq-artemis/pull/2522#issuecomment-457819129
@michaelandrepearce
> Have you got any stats to share on this, especially on running on
standardish spec server? Im assuming that its quite a good improvement based on
it removes copies :) , but be good for the record like....
Not yet: I can just say that on my box `fsync` is such heavy-weight that is
difficult to measure the gain of this improvement.
If I would use a `tmpfs` folder for the journal instead it would just show
that zeroing (and copying) is just the major cost and the speedup is so
dramatical that is just unrealistic.
@clebertsuconic @michaelandrepearce I will try to get a proper benchmark box
to find out this, but nonetheless I'm expecting *at least* a better ability
from NIO/MAPPED to maintain the buffer timeout: the next big improvement would
be to run `fsync` on a separate "coalescing" thread (similarly to what libaio
does on the blocked poll).
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 190378)
Time Spent: 4.5h (was: 4h 20m)
> Zero-copy NIO/MAPPED TimedBuffer
> --------------------------------
>
> Key: ARTEMIS-2239
> URL: https://issues.apache.org/jira/browse/ARTEMIS-2239
> Project: ActiveMQ Artemis
> Issue Type: Improvement
> Components: Broker
> Affects Versions: 2.6.4
> Reporter: Francesco Nigro
> Assignee: Francesco Nigro
> Priority: Major
> Time Spent: 4.5h
> Remaining Estimate: 0h
>
> NIO/MAPPED journal types are currently performing the copy of the buffer used
> on TimedBuffer to accumulate writes, but this copy (and the zeroing of the
> new buffer) could be avoided, reducing dramatically the CPU usage and
> latencies during a burst of writes.
> An added benefit of this change is that isn't necessary anymore to stress the
> thread pool of the file factory, because there is no need to pool the buffer
> used to perform the write.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
