adriangb commented on PR #21182: URL: https://github.com/apache/datafusion/pull/21182#issuecomment-4190429069
> 8MB was too small for wide-row full scans (Q3: SELECT * with 16 columns), causing SPM to stall on I/O. 64MB per partition is still strictly less than the SortExec it replaces (which buffers entire partition in memory). BufferExec integrates with MemoryPool so it won't cause OOM. I think this is right. Basically: `SortExec` is "unlimited" buffering. IMO we could go even higher if we have to a pick a number (although perhaps it should be configurable if it isn't already) - something like 512MB. If the partition is smaller it will never be hit. If it is larger or we run out of memory it will spill. But let's see what the numbers look like with 64MB. -- 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]
