[
https://issues.apache.org/jira/browse/IGNITE-13886?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17253988#comment-17253988
]
Nikolay Izhikov commented on IGNITE-13886:
------------------------------------------
[~dmekhanikov]
I'm +1 to unify all measure units.
What precision do we need?
Is it enough to have millisecond precision?
{{U.currentTimeMillis()}} we use works very unexpectedly and gave wrong results
for clear measures.
I think we should get rid of the current implementation
{{U.currentTimeMillis()}} in the first place and then migrate to millisecond
precision.
> Change units of cache-related histograms to milliseconds
> --------------------------------------------------------
>
> Key: IGNITE-13886
> URL: https://issues.apache.org/jira/browse/IGNITE-13886
> Project: Ignite
> Issue Type: Improvement
> Reporter: Denis Mekhanikov
> Priority: Major
>
> Ignite has different metrics that have the "histogram" type:
> * tx.nodeSystemTimeHistogram
> * tx.nodeUserTimeHistogram
> * pme.DurationHistogram
> * pme.CacheOperationsBlockedDurationHistogram
> * cache.<cacheName>.GetTime
> * cache.<cacheName>.PutTime
> * cache.<cacheName>.RemoveTime
> * cache.<cacheName>.CommitTime
> * cache.<cacheName>.RollbackTime
> First four have buckets corresponding to the amount of time the operation
> took in milliseconds.
> Cache-related histograms are measured in nanoseconds, while it would be
> enough to use milliseconds there as well.
> It's hard to distinguish between 100000000 and 1000000000 nanoseconds
> visually.
> The following set of buckets should be used:
> * 1
> * 10
> * 100
> * 250
> * 1000
> The values are provided in milliseconds.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
