You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ozone.apache.org by "Sammi Chen (Jira)" <ji...@apache.org> on 2021/09/28 08:49:00 UTC

[jira] [Updated] (HDDS-5791) Use Guava Cache to replace current ResourceLimitCache for stateMachineDataCache

     [ https://issues.apache.org/jira/browse/HDDS-5791?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Sammi Chen updated HDDS-5791:
-----------------------------
    Description: 
In the current ResourceLimitCache, write request data is removed from the stateMachineDataCache when the majority of the ratis peers have committed the request, which leaves the leader to fetch the write data from ContainerStateMachine if it hasn't assemble and send the corresponding appendEntities to the left follower. 

This task amis to improve the stateMachineDataCache hit rate by lazily remove element, that is, keep the element in the cache as long as possilbe. 


One drawback of using Gauva Cache is it cannot control the total cased data size, only data count is controlable. 

  was:
In the current ResourceLimitCache, write request data is removed from the stateMachineDataCache when the majority of the ratis peers have committed the request, which leaves the leader to fetch the write data from ContainerStateMachine if it hasn't assemble and send the corresponding appendEntities to the left follower. 

This task amis to improve the stateMachineDataCache hit rate by lazily remove element, that is, keep the element in the cache as long as possilbe. 


> Use Guava Cache to replace current ResourceLimitCache for stateMachineDataCache
> -------------------------------------------------------------------------------
>
>                 Key: HDDS-5791
>                 URL: https://issues.apache.org/jira/browse/HDDS-5791
>             Project: Apache Ozone
>          Issue Type: Improvement
>            Reporter: Sammi Chen
>            Assignee: Sammi Chen
>            Priority: Major
>
> In the current ResourceLimitCache, write request data is removed from the stateMachineDataCache when the majority of the ratis peers have committed the request, which leaves the leader to fetch the write data from ContainerStateMachine if it hasn't assemble and send the corresponding appendEntities to the left follower. 
> This task amis to improve the stateMachineDataCache hit rate by lazily remove element, that is, keep the element in the cache as long as possilbe. 
> One drawback of using Gauva Cache is it cannot control the total cased data size, only data count is controlable. 



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@ozone.apache.org
For additional commands, e-mail: issues-help@ozone.apache.org