[IGNITE-10876] "Affinity changes (coordinator) applied" can be executed in parallel

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 2.8
• Assignee: Pavel Voronkin
• Reporter: Pavel Voronkin
• Created: 10/Jan/19 06:29
• Updated: 04/Feb/19 11:47
• Resolved: 04/Feb/19 11:45

Description

There is for loop over all cache groups which execution N*P operations in exchange worker where N is number of cache groups, P is number of partitions.

We spend 80% of time in a loop

for (CacheGroupContext grp : cctx.cache().cacheGroups())

{ GridDhtPartitionTopology top = grp != null ? grp.topology() : cctx.exchange().clientTopology(grp.groupId(), events().discoveryCache()); top.beforeExchange(this, true, true); }

 

I believe we can execute it in parallel

Issue Links

• links to: GitHub Pull Request #5942
• links to: PR