[ 
https://issues.apache.org/jira/browse/CASSANDRA-6591?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13922755#comment-13922755
 ] 

Ian Barfield commented on CASSANDRA-6591:
-----------------------------------------

The "hit mva / rate mva" and "hit rate mva" as you define them are indeed very 
different things. This is most likely intentional and I would say desirable. 
The red line "hit rate mva" seems to be applying the idea of recency to a 
metric that already has a recency window (in this case the window is 'all 
time'). This strikes me as both unintuitive and unhelpful; after the metrics 
have been going for a while it is unlikely to change very fast, and by making 
it a moving average it will change even slower. You could easily have a 100% 
miss ratio for several minutes and never see that line move. That seems to 
defeat the purpose of having a "recent" hitRate metric.

As for EWMAs being 'comparable': I'm not certain on the exact mathematical 
implications of dividing two estimates after sampling, but I strongly suspect 
it would be more than accurate enough for this purpose.

> un-deprecate cache recentHitRate and expose in o.a.c.metrics
> ------------------------------------------------------------
>
>                 Key: CASSANDRA-6591
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-6591
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>            Reporter: Chris Burroughs
>            Assignee: Chris Burroughs
>            Priority: Minor
>         Attachments: j6591-1.2-v1.txt, j6591-1.2-v2.txt, j6591-1.2-v3.txt
>
>
> recentHitRate metrics were not added as part of CASSANDRA-4009 because there 
> is not an obvious way to do it with the Metrics library.  Instead hitRate was 
> added as an all time measurement since node restart.
> This does allow changes in cache rate (aka production performance problems)  
> to be detected.  Ideally there would be 1/5/15 moving averages for the hit 
> rate, but I'm not sure how to calculate that.  Instead I propose updating 
> recentHitRate on a fixed interval and exposing that as a Gauge.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Reply via email to