[ 
https://issues.apache.org/jira/browse/KAFKA-13188?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17398148#comment-17398148
 ] 
Alok Nikhil commented on KAFKA-13188:
-------------------------------------

[~luwang] I see that the consumer side memory pool management KIP was never 
implemented - 
[https://cwiki.apache.org/confluence/display/KAFKA/KIP-81:+Bound+Fetch+memory+usage+in+the+consumer]

The default is `MemoryPool.NONE`. How did LinkedIn measure the performance 
here? Is there a customization / config that allows the consumer to use a 
different memory pool?

> Release the memory back into MemoryPool
> ---------------------------------------
>
>                 Key: KAFKA-13188
>                 URL: https://issues.apache.org/jira/browse/KAFKA-13188
>             Project: Kafka
>          Issue Type: Improvement
>            Reporter: Lucas Wang
>            Assignee: Alok Nikhil
>            Priority: Major
>             Fix For: 3.0.1
>
>
> Tushar made a [hotfix change|https://github.com/linkedin/kafka/pull/186] to 
> the linkedin/kafka repo hosting apache kafka 2.4.
> The change is about releasing memory back to the MemoryPool for the kafka 
> consumer, and his benchmark showed significant improvement in terms of the 
> memory graduating from Young Gen and promoted to Old Gen.
> Given the benefit, the change should also be added trunk.



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

Reply via email to