Dandandan commented on code in PR #2308: URL: https://github.com/apache/datafusion-ballista/pull/2308#discussion_r3789962687
########## docs/source/contributors-guide/architecture.md: ########## @@ -194,6 +194,10 @@ Each executor will re-partition the output of the stage it is running so that it stage. This mechanism is known as an Exchange or a Shuffle. The logic for this can be found in the [ShuffleWriterExec] and [ShuffleReaderExec] operators. +The shuffle is _blocking_: a stage runs to completion and materializes its output to local disk before any Review Comment: In principle some next stages can start when partitions are written (without waiting on the full stage) - currently it waits on the full stage but in some cases it can beneficial (when enough cluster capacity) to already start the next stage... -- 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]
