You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Pavel Kovalenko (JIRA)" <ji...@apache.org> on 2018/08/14 17:09:00 UTC
[jira] [Updated] (IGNITE-9270) Design thread per partition model
[ https://issues.apache.org/jira/browse/IGNITE-9270?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Pavel Kovalenko updated IGNITE-9270:
------------------------------------
Description:
A new model of executing cache partition operations (READ, CREATE, UPDATE, DELETE) should satisfy following conditions
1) All modify operations (CREATE, UPDATE, DELETE) on some partition must be performed by the same thread.
2) Read operations can be executed by any thread.
3) Ordering of modify operations on primary and backup nodes should be same.
We should investigate performance if we choose dedicated executor service for such operations, or we can use a messaging model to use network threads to perform such operations.
was:
A new model of executing cache partition operations (READ, CREATE, UPDATE, DELETE) should satisfy following conditions
1) All modify operations (CREATE, UPDATE, DELETE) on some partition must be performed by the same thread.
2) Read operations can be executed by any thread.
We should investigate performance if we choose dedicated executor service for such operations, or we can use a messaging model to use network threads to perform such operations.
> Design thread per partition model
> ---------------------------------
>
> Key: IGNITE-9270
> URL: https://issues.apache.org/jira/browse/IGNITE-9270
> Project: Ignite
> Issue Type: Sub-task
> Components: cache
> Reporter: Pavel Kovalenko
> Assignee: Pavel Kovalenko
> Priority: Major
> Fix For: 2.7
>
>
> A new model of executing cache partition operations (READ, CREATE, UPDATE, DELETE) should satisfy following conditions
> 1) All modify operations (CREATE, UPDATE, DELETE) on some partition must be performed by the same thread.
> 2) Read operations can be executed by any thread.
> 3) Ordering of modify operations on primary and backup nodes should be same.
> We should investigate performance if we choose dedicated executor service for such operations, or we can use a messaging model to use network threads to perform such operations.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)