hiboyang opened a new pull request #32031:
URL: https://github.com/apache/spark/pull/32031


   ### What changes were proposed in this pull request?
   
   This PR contains Remote Shuffle Service to support dynamic allocation on 
Kubernetes. The code is mostly copied from [Uber Remote Shuffle 
Service](https://github.com/uber/RemoteShuffleService) and modified with some 
renaming. Also added Kubernetes related support which does not exist in 
original Uber Remote Shuffle Service.
   
   ### Why are the changes needed?
   
   It is still difficult to use dynamic allocation with Spark on Kubernetes. 
There are several disaggregated/remote shuffle solutions in different 
companies. Hopefully we could get a remote shuffle implementation into Spark 
and enhanced in the future by the Spark community.
   
   ### Does this PR introduce _any_ user-facing change?
   
   Yes, user could set Spark config 
(spark.shuffle.manager=org.apache.spark.shuffle.RssShuffleManager) to run Spark 
applications with remote shuffle service. It will make Spark use the new 
RssShuffleManager to write/read shuffle data to/from remote shuffle service.
   
   ### How was this patch tested?
   
   Manually tested with Spark application in Kubernetes.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
[email protected]



---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to