[
https://issues.apache.org/jira/browse/ARTEMIS-1476?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16219631#comment-16219631
]
ASF GitHub Bot commented on ARTEMIS-1476:
-----------------------------------------
Github user franz1981 commented on the issue:
https://github.com/apache/activemq-artemis/pull/1605
> It doesn't really matter to add histograms at the sync-calculation, or to
perform any writes...
That's the part I'm trying to understand: IMHO it matters to have a more
detailed view of the sync test using standard de facto tools too as
@michaelandrepearce has stated above in other comments.
While having it in production in real time (as another PR I've already
pushed has provided) is a different but equally (maybe more) important purpose.
With IO, treating the latency just as 1/throughput risks to hide important
informations: providing a test that shows the distribution of the percentiles
could give to the expert user the chance to tune the buffer timeout based on
the 99 percentile (taken from the real hw) instead of an average in order to be
sure to met his SLA.
> HdrHistogram support on verbose SyncCalculation
> -----------------------------------------------
>
> Key: ARTEMIS-1476
> URL: https://issues.apache.org/jira/browse/ARTEMIS-1476
> Project: ActiveMQ Artemis
> Issue Type: Improvement
> Components: Broker
> Reporter: Francesco Nigro
> Assignee: Francesco Nigro
>
> The SyncCalculation::syncTest could benefit of
> [HdrHistogram|http://hdrhistogram.org/] support to represent the write
> latencies percentile distribution while using it as a CLI command (ie through
> the perf-journal command).
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
