Re: Range Merge Join v1

From: Andrew Borodin <borodin(at)octonica(dot)com>
To: Jeff Davis <pgsql(at)j-davis(dot)com>
Cc: PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Range Merge Join v1
Date: 2017-04-21 17:15:07
Message-ID: CAJEAwVFrvc_ude4=Rv53x1AM5-nZro0jEo2XydHMrA1A+bRyqw@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox
Thread:
Lists: pgsql-hackers

Hi, Jeff!

I'm planning to provide the review for this patch in future. I've done
some performance tesing (see attachment).
All in all, nothing important, everything works fine.
Tests were executed under current HEAD on Ubuntu 16 over MacBook Air.

I observe that:
1. Patch brings performance improvement for specified query
2. Performance improvement of Range Merge Join vs GiST Nested Loop
Join (on Index Only Scan) is between 4x and 6x
3. Performance improvement of RMJ with B-tree index vs GiST is between
4x and 10x
(due to lack of actually competing cases, here I omit T-tests, they
are not that important when speaking about 4x difference)
4. Range Merge Join is capable to consume expressional index with
exactly same effect as direct index
5. Other attributes residing in joined tables do not affect
performance improvement

I'll make code review some time later, probably next week. (I hope
there is no urge in this, since commitfest is in July, or is it
urgent?) BTW fixe typo in index specification of original performance
test (both indexes were for same table)

Best regards, Andrey Borodin.

Attachment Content-Type Size
perf2.sql application/octet-stream 7.4 KB

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Andres Freund 2017-04-21 17:20:44 Re: Unportable implementation of background worker start
Previous Message Tom Lane 2017-04-21 16:50:04 Re: Unportable implementation of background worker start