You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Ilya Lantukh (JIRA)" <ji...@apache.org> on 2017/08/18 16:30:00 UTC
[jira] [Commented] (IGNITE-5872) Replace standard java maps for
partition counters with more effective data structures
[ https://issues.apache.org/jira/browse/IGNITE-5872?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16133218#comment-16133218 ]
Ilya Lantukh commented on IGNITE-5872:
--------------------------------------
I've performed code review and found a few minor issues:
- GridDiscoveryManager.createDiscoCache(...) - why shouldn't we do minProdVer = loc.version() in the beginning? Otherwise it might be possible in certain corner cases that DiscoCache.minProdVer == null.
- SchemaIndexCacheVisitorImpl.processPartition(...) - looks like it was modified to fix another, independent issue. I think it need further investigation because fix looks very strange.
And, in general, I would like to see these optimized maps in continuous queries as well.
> Replace standard java maps for partition counters with more effective data structures
> -------------------------------------------------------------------------------------
>
> Key: IGNITE-5872
> URL: https://issues.apache.org/jira/browse/IGNITE-5872
> Project: Ignite
> Issue Type: Improvement
> Components: cache
> Affects Versions: 2.1
> Reporter: Alexey Goncharuk
> Assignee: Alexey Goncharuk
> Fix For: 2.2
>
>
> Currently we have map of Integer -> Tuple <Long, Long> which is passed for each cache group. This is very inefficient as at causes a lot of boxes and overheads in marshalling.
> This particular map can be replaced with a sorted primitive array.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)