You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Semen Boikov (JIRA)" <ji...@apache.org> on 2016/06/08 12:33:20 UTC

[jira] [Assigned] (IGNITE-3261) AffinityKey is not stored in the metadata cache

     [ https://issues.apache.org/jira/browse/IGNITE-3261?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Semen Boikov reassigned IGNITE-3261:
------------------------------------

    Assignee: Semen Boikov

> AffinityKey is not stored in the metadata cache
> -----------------------------------------------
>
>                 Key: IGNITE-3261
>                 URL: https://issues.apache.org/jira/browse/IGNITE-3261
>             Project: Ignite
>          Issue Type: Bug
>          Components: cache
>    Affects Versions: 1.5.0.final
>            Reporter: Denis Magda
>            Assignee: Semen Boikov
>            Priority: Critical
>              Labels: community, important
>             Fix For: 1.7
>
>         Attachments: Ignite-MarshallBenchmark.zip
>
>
> Presently we don't register predefined and system classes in metadata cache which can lead to significant performance drops when these types used as keys.
> As an example we have {{AffinityKey}} class. It's not registered in the metadata cache and as a result client nodes don't update their {{CacheObjectBinaryProcessorImpl.clientMetaDataCache}}. After that when a client node needs to get {{AffinityKey}} metadata using {{CacheObjectBinaryProcessorImpl.metadata(typeId)}} it will always call metadata cache and this is a bottleneck. The drop can be significant because this method is called from methods like {{GridAffinityProcessor.mapKeyToPrimaryAndBackups}}.
> In attach you can find a simple benchmark that shows how slower a result if AffinityKey is used.
> As a solution we can register {{AffinityKey}} and other system and predefined classes (?) in the metadata cache.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)