[
https://issues.apache.org/jira/browse/HBASE-9806?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14002238#comment-14002238
]
stack commented on HBASE-9806:
------------------------------
Should this be in the io.hfile package since that is where BC's normally fester.
PE has more than PerfEval at the moment (histograms, rows by data size). This
does gaussian rows which PE doesn't have which would be nice (but for
BlockCache testing we want random-sized data I'd say). Should this go in
[~ndimiduk]? Or want me to adopt to PE? Add a random data size? Will help
yeah w/ looking at BC options quickly on one machine.
> Add PerfEval tool for BlockCache
> --------------------------------
>
> Key: HBASE-9806
> URL: https://issues.apache.org/jira/browse/HBASE-9806
> Project: HBase
> Issue Type: Test
> Components: Performance, test
> Reporter: Nick Dimiduk
> Assignee: Nick Dimiduk
> Attachments: HBASE-9806.00.patch, HBASE-9806.01.patch,
> HBASE-9806.02.patch, conf_20g.patch, conf_3g.patch, test1_run1_20g.pdf,
> test1_run1_3g.pdf, test1_run2_20g.pdf, test1_run2_3g.pdf
>
>
> We have at least three different block caching layers with myriad
> configuration settings. Let's add a tool for evaluating memory allocations
> and configuration combinations with different access patterns.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
