Обсуждение: [FWD] About "Our CLUSTER implementation is pessimal" patch
I really thought this would have caused some interest, since - this item is in the TODO list - the improvement for CLUSTER in some scenarios is 800%, and maybe more (if I didn't do anything wrong, of course...) Could at least the message: http://archives.postgresql.org/pgsql-hackers/2010-02/msg00766.php be added to the TODO page, under "Improve CLUSTER performance by sorting to reduce random I/O" ? It would be sad if the patch got lost... Leonardo > Attached the updated patch (should solve a bug) and a script. > The sql scripts generates a 2M rows table ("orig"); then the > table is copied and the copy clustered using seq + sort (since > "set enable_seqscan=false;"). > Then the table "orig" is copied again, and the copy clustered > using regular index scan (set enable_indexscan=true; set > enable_seqscan=false). > Then the same thing is done on a 5M rows table, and on a 10M > rows table. > > On my system (Sol10 on a dual Opteron 2.8) single disc: > > > 2M: seq+sort 11secs; regular index scan: 33secs > 5M: seq+sort 39secs; regular index scan: 105secs > 10M:seq+sort 83secs; regular index scan: 646secs > > > Maybe someone could suggest a better/different test? > > > Leonardo
Leonardo F wrote: > Could at least the message: > http://archives.postgresql.org/pgsql-hackers/2010-02/msg00766.php > be added to the TODO page, under > "Improve CLUSTER performance by sorting to reduce > random I/O" ? > It would be sad if the patch got lost... You should read http://wiki.postgresql.org/wiki/Submitting_a_Patch and follow the instructions there. As outlined in the "Submission timing" section, you're asking about something during the wrong time to be doing so--that's why you're not getting any real feedback. Add your patch to the next CommitFest by linking to your message at https://commitfest.postgresql.org/ and it won't get lost--it will get assigned a reviewer at the time when the time comes to look at completely new patches again. -- Greg Smith 2ndQuadrant Baltimore, MD PostgreSQL Training, Services and Support greg@2ndQuadrant.com www.2ndQuadrant.com
> As outlined in the "Submission timing" section, you're > asking about something during the wrong time to be doing so--that's why you're > not getting any real feedback. Add your patch to the next CommitFest by linking > to your message at https://commitfest.postgresql.org/ Ok! But there's something I don't understand: I didn't add the patch to the next CommitFest because I thought it could never be added in 9.0 (because it adds a new "feature" which has never been discussed). Hence I thought it should have been "discussed" (not properly "reviewed") out of a CommitFest. The "Submission timing" section talks about "beta phase", not "alpha phase", so I'm stll confused... In other words: should patches that won't be included in the next release (because it's too late) still added to the next CommitFest? I thought a very "rough" discussion was the way to go in these cases, but I'm not familiar at all with the process... I'll wait for an answer before adding the patch to the CommitFest (and in case, I'll add more comments and docs to it) Thank you very much! Leonardo
Leonardo F wrote: > But there's something I don't understand: I didn't add the patch to the next > CommitFest because I thought it could never be added in 9.0 (because it adds a > new "feature" which has never been discussed). Hence I thought it should have > been "discussed" (not properly "reviewed") out of a CommitFest. > The "Submission timing" section talks about "beta phase", not "alpha phase", so > I'm stll confused... > In other words: should patches that won't be included in the next release > (because it's too late) still added to the next CommitFest? Yes. There's not going to be any more commitfests for this release, so the next commitfest is for 9.1. (don't worry about the lack of enthusiasm for the patch, people are just very busy with 9.0 and don't have the energy to think about 9.1 material at this point) -- Heikki Linnakangas EnterpriseDB http://www.enterprisedb.com
> Yes. There's not going to be any more commitfests for this release, so > the next commitfest is for 9.1. Perfect! Where could I find such information? I mean: how could I know it? > (don't worry about the lack of enthusiasm for the patch, people are just > very busy with 9.0 and don't have the energy to think about 9.1 material > at this point) I understand! Since it's going to be posted in a CommitFest, I'll try to clean up the patch a little. Thank you Greg and Heikki. Leonardo