Log In - Home Page

Patch: Revise parallel pg_restore's scheduling heuristic

Edit Patch - Move To Another CommitFest - Delete Patch

CommitFest 2009-07
Topic Performance
Patch Status Committed
Author Tom Lane
Reviewers Kevin Grittner
Committer Nobody
Close Date 2009-08-07
Comment by tgl on 2009-05-15 06:37:25 PM: There is only a theoretical argument that this might be better than what's there now; we need testing to see if the theory is right.
Patch by tgl on 2009-05-29 09:42:53 PM: Initial version
Patch by kgrittn on 2009-07-16 08:57:41 PM: Version 2 rebases for pg_indent
Preliminary checks OK. Performance tests to follow.
Comment by rhaas on 2009-07-19 05:36:40 AM: Kevin needs another week to test this.
Review by rhaas on 2009-07-27 10:31:59 AM: Kevin's initial tests show no difference; why?
Review by rhaas on 2009-07-30 11:39:00 AM: Kevin finds a small benefit.
Patch by tgl on 2009-07-31 09:34:36 AM: v3, this time *really* without O(N^2) behavior
Comment by kgrittn on 2009-07-31 03:44:46 PM: daveg may be able to benchmark.
Comment by kgrittn on 2009-07-31 03:46:10 PM: Kevin will benchmark new patch.
Comment by rhaas on 2009-08-02 09:42:57 AM: Seems there is enough evidence to commit this even while followup research is continuing.
Comment by tgl on 2009-08-05 11:45:55 AM: I'm discouraged by the complete lack of evidence of benefit. Maybe we should just abandon this.
Comment by rhaas on 2009-08-05 07:21:28 PM: Should be discussed on -hackers, but I share your concern.

Add Comment

Please log in to comment on this patch.