Shekharrajak opened a new pull request, #3903:
URL: https://github.com/apache/datafusion-comet/pull/3903

   ## Which issue does this PR close?
   
   https://github.com/apache/datafusion-comet/issues/3859
   
   ## Rationale for this change
   
   Before: N partitions x S spills = up to N*S temp files (e.g., 200 partitions 
x 5 spills = 1000 files)
   After: S temp files total (e.g., 5 files regardless of partition count)
   Each saved file avoids open() + write() + close() + open() + 
copy_file_range() syscalls
   On EBS-backed storage: each metadata op costs 1-5ms, so saving ~995 ops can 
save 1-5 seconds per task
   
   ## What changes are included in this PR?
   
   Rewrite spill.rs -- Replace PartitionWriter with a PartitionWriter that no 
longer owns spill files, plus new SpillInfo / PartitionSpillRange types
   
   Rewrite spill/merge in multi_partition.rs -- Single spill file per event, 
seek-based merge
   
   
   ## How are these changes tested?
   
   Update tests in shuffle_writer.rs -- Adapt existing test, add new ones


-- 
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]

Reply via email to