Csaba Ringhofer has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/22217 )

Change subject: WIP IMPALA-13475: Mem limit deferred RPC processing parallelism
......................................................................


Patch Set 8:

> memory to process. I think the current way we do this can produce
> starvation for large deferred messages, and we absolutely should do

Good point, I kind of thought about it, but didn't realize how bad this can 
become - with a very slow consumer there may be only space for always for one 
small batch, delaying others for a very long time.

> changes to push the total_buffer_limit_ and num_buffered_bytes_ out
> to the SenderQueue to avoid the CAS might be a good thing, but they
> also seem a bit orthogonal to the starvation problem.

This part is actually needed for merging exchanges to track their quote 
individually. Added a section about this to the commit message.


--
To view, visit http://gerrit.cloudera.org:8080/22217
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I4d8e2b07a19e4d30210d7903c01e75fba632a1de
Gerrit-Change-Number: 22217
Gerrit-PatchSet: 8
Gerrit-Owner: Csaba Ringhofer <[email protected]>
Gerrit-Reviewer: Csaba Ringhofer <[email protected]>
Gerrit-Reviewer: Impala Public Jenkins <[email protected]>
Gerrit-Reviewer: Joe McDonnell <[email protected]>
Gerrit-Comment-Date: Mon, 23 Mar 2026 20:44:50 +0000
Gerrit-HasComments: No

Reply via email to