[
https://issues.apache.org/jira/browse/IGNITE-9561?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16631612#comment-16631612
]
Pavel Kovalenko commented on IGNITE-9561:
-----------------------------------------
[~Pavlukhin] Well, it mostly depends on the number of cache groups and
partitions per each of cache group. I observe 40-50 seconds affinity
calculation in a single thread on some of the production environment with a lot
of a number of cache groups and partitions.
> Optimize affinity initialization for started cache groups
> ---------------------------------------------------------
>
> Key: IGNITE-9561
> URL: https://issues.apache.org/jira/browse/IGNITE-9561
> Project: Ignite
> Issue Type: Improvement
> Components: cache
> Affects Versions: 2.5
> Reporter: Pavel Kovalenko
> Assignee: Pavel Kovalenko
> Priority: Major
> Fix For: 2.8
>
>
> At the end of
> {noformat}
> org.apache.ignite.internal.processors.cache.CacheAffinitySharedManager#processCacheStartRequests
>
> {noformat}
> method we're initializing affinity for cache groups starting at current
> exchange.
> We do it one-by-one and synchronously wait for AffinityFetchResponse for each
> of the starting groups. This is inefficient. We may parallelize this process
> and speed up caches starting process.
> NOTE: There are also a lot of affinity recalculation methods in:
> {noformat}
> CacheAffinitySharedManager
> {noformat}
> which all looks like iterate over cache groups and recalculate affinity for
> all of them. We can easily speed-up each of such methods executing in
> parallel affinity re-calculation for each of cache groups.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
