[
https://issues.apache.org/jira/browse/IGNITE-3018?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15279802#comment-15279802
]
Taras Ledkov commented on IGNITE-3018:
--------------------------------------
Unfortunately bucket based distribution has a very bad re-balance index.
In the test below the value is the average partitions count that are changed
after each change of the topology.
Test 60 nodes. Old: 11.4; New: 19.8;
Test 100 nodes. Old: 7.6; New: 135.7;
Test 200 nodes. Old: 4.0; New: 100.1;
Test 300 nodes. Old: 2.7; New: 75.1;
Test 400 nodes. Old: 1.9; New: 60.5;
Test 500 nodes. Old: 1.9; New: 52.5;
Test 600 nodes. Old: 1.2; New: 46.0;
So, it looks like we have to use simple sort all hash(partition, node) to
prevent rebalance problem.
> Cache affinity calculation is slow with large nodes number
> ----------------------------------------------------------
>
> Key: IGNITE-3018
> URL: https://issues.apache.org/jira/browse/IGNITE-3018
> Project: Ignite
> Issue Type: Bug
> Components: cache
> Reporter: Semen Boikov
> Assignee: Taras Ledkov
> Priority: Critical
> Fix For: 1.7
>
> Attachments: 003.png, 064.png, 100.png, 128.png, 200.png, 300.png,
> 400.png, 500.png, 600.png
>
>
> With large number of cache server nodes (> 200) RendezvousAffinityFunction
> and FairAffinityFunction work pretty slow .
> For RendezvousAffinityFunction.assignPartitions can take hundredes of
> milliseconds, for FairAffinityFunction it can take seconds.
> For RendezvousAffinityFunction most time is spent in MD5 hash calculation and
> nodes list sorting. As optimization we can try to cache {partion, node} MD5
> hash or try another hash function. Also several minor optimizations are
> possible (avoid unncecessary allocations, only one thread local 'get', etc).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
