[
https://issues.apache.org/jira/browse/HBASE-12133?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14154445#comment-14154445
]
Yi Deng commented on HBASE-12133:
---------------------------------
[~stack] Thanks for commenting.
We're going to use it to compute a P99/P95 metric every one minute. The whole
design is to make a none-synchronized version of histogram collecting.
About the updateMin/updateMax, the implementation is same as what the blog
shows and it is good for me. People there are mainly concerning about whether
that should be part of JDK.
In practice, even if a lot of threads are calling the same method, all of them
could finish without many retrying, because the retry only happened if
1) the value needs to be change (i.e. new value is smaller than the current
minimum). Actually, most of the thread will not try to change the value at all.
2) some other thread change the current value(not just execute parralelly),
AFTER the current value is taken out and BEFORE the new value is sucessfully
compared and set. Since the code here is so fast that we can be optimistic that
it only happens with a very small probabilty. (but we don't assume it never
happen)
3) since the value is changing to a single direction (keep decreasing for
updateMin), it is very likely in the next retry (because the value has been
changed smaller), the updated value has been samller or equal to the value to
be set to, and then quit the loop.
> Add FastLongHistogram for metric computation
> --------------------------------------------
>
> Key: HBASE-12133
> URL: https://issues.apache.org/jira/browse/HBASE-12133
> Project: HBase
> Issue Type: New Feature
> Components: metrics
> Affects Versions: 0.98.8
> Reporter: Yi Deng
> Assignee: Yi Deng
> Priority: Minor
> Labels: histogram, metrics
> Fix For: 0.98.8
>
> Attachments:
> 0001-Add-FastLongHistogram-for-fast-histogram-estimation.patch
>
>
> FastLongHistogram is a thread-safe class that estimate distribution of data
> and computes the quantiles. It's useful for computing aggregated metrics like
> P99/P95.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
