xianjingfeng opened a new issue, #2107: URL: https://github.com/apache/incubator-uniffle/issues/2107
### Code of Conduct - [X] I agree to follow this project's [Code of Conduct](https://www.apache.org/foundation/policies/conduct) ### Search before asking - [X] I have searched in the [issues](https://github.com/apache/incubator-uniffle/issues?q=is%3Aissue) and found no similar issues. ### What would you like to be improved? Currently, Uniffle client will create an `EventLoopGroup` for each shuffle server, which is named `workerGroup`. And the default number of the threads of `workerGroup` is `availableProcessors` and it can be changed by `rss.client.netty.client.threads`. And each thread of `workerGroup` will create an `EventExecutor`, which will occur lots of memory. https://github.com/apache/incubator-uniffle/blob/d170004a32418a32536aa792b198dfd714b5ea8c/common/src/main/java/org/apache/uniffle/common/netty/client/TransportClientFactory.java#L88 But the param `rss.client.netty.client.connections.per.peer`, which with 2 as the default value, will de determine the max threads in `workerGroup` will be used. So, many threads are useless in the default configuration. https://github.com/apache/incubator-uniffle/blob/d170004a32418a32536aa792b198dfd714b5ea8c/common/src/main/java/org/apache/uniffle/common/netty/client/TransportClientFactory.java#L127-L133 ### How should we improve? Remove `rss.client.netty.client.threads` and make `rss.client.netty.client.connections.per.peer` as the param of the thread number of `workerGroup`. ### Are you willing to submit PR? - [X] Yes I am willing to submit a PR! -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
