tag:blogger.com,1999:blog-20038672.post6330562123412755964..comments2024-03-28T00:58:29.187-04:00Comments on Robert Haas: Rob Wultsch's MySQL Talk at PostgreSQL WestRobert Haashttp://www.blogger.com/profile/08393677427643988650noreply@blogger.comBlogger10125tag:blogger.com,1999:blog-20038672.post-76878189778865375252010-11-15T23:01:12.786-05:002010-11-15T23:01:12.786-05:00href="http://blogs.innodb.com/wp/2010/09/mysq...<a rel="nofollow">href="http://blogs.innodb.com/wp/2010/09/mysql-5-5-innodb-change-buffering/">This describes</a> how index-only queries are supported for InnoDB. Details are sparse. Maybe Postgres can use the same approach?<br /><br />I have had more fun with the MySQL optimizer than Rob has. Fortunately query plan hints fix most problems and I don't deal with complex queries/SQL.<br /><br />I hope to see the Postgres crowd at the OReilly conference next year. I also hope that someone does a Postgres for advanced/intermediate MySQL users during that week. I will attend that.Mark Callaghanhttps://www.blogger.com/profile/09590445221922043181noreply@blogger.comtag:blogger.com,1999:blog-20038672.post-6309948267648706012010-11-14T10:35:12.330-05:002010-11-14T10:35:12.330-05:00@Baron:
I agree in general with query cache being ...@Baron:<br />I agree in general with query cache being something that MySQL does not do well. For all of its issues, the query cache is incredibly helpful for certain workloads.<br /><br />Adding memcached is another layer of complexity but usually the right choice.Robhttps://www.blogger.com/profile/13656862909374634867noreply@blogger.comtag:blogger.com,1999:blog-20038672.post-11276981532011473642010-11-12T18:52:22.540-05:002010-11-12T18:52:22.540-05:007. Cheap connections. IMHO, it is something we sho...7. Cheap connections. IMHO, it is something we should invest some time. Almost every DBMS has an integrated pooler solution. We're resisting to put some solution "near" the backend code (At least <a href="http://archives.postgresql.org/pgsql-hackers/2009-12/msg02192.php" rel="nofollow">someone</a> already made a proposal.) We already learned with all of the replication stuff.<br /><br />I'm advocating to write an standard pooler for PostgreSQL. We already have some solid knowledge about PostgreSQL poolers for now. Of course, the existent poolers could be used (pooler_mode = off).Euler Taveirahttps://www.blogger.com/profile/08335071854403271396noreply@blogger.comtag:blogger.com,1999:blog-20038672.post-14670247231363316922010-11-11T18:52:04.421-05:002010-11-11T18:52:04.421-05:00As much as I prefer PostgreSQL over MySQL, I have ...As much as I prefer PostgreSQL over MySQL, I have to agree that VACUUM is probably the most annoying thing to deal with in Postgres. It did get better with 8.3 and 8.4 but it is still a major headache compared to other DBMS (not only MySQL).<br /><br />As far as compression is concerned: I find DB2's dictionary basec compression pretty cool and it apparently gives pretty good compression ratios.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-20038672.post-44743406033398032322010-11-11T09:20:24.491-05:002010-11-11T09:20:24.491-05:00Nigel Tufnel thinks you should set default_statist...Nigel Tufnel thinks you should set default_statistics_target to 11, to give Postgres that extra push when you need it.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-20038672.post-43240310800294305352010-11-11T08:49:07.222-05:002010-11-11T08:49:07.222-05:005. ... The benefit of this is that we can abort t...5. ... The benefit of this is that we can abort transactions very, very quickly, and we can essentially forget about them as fast as we abort them. InnoDB must keep aborted transaction IDs in each MVCC snapshot until they've been cleaned up.<br /><br />The issue here boils down to application semantics: do they routinely rollback? If not, then the MySql approach (call 'optimistic') gives better performance.<br /><br /><br />8. Query cache.<br /><br />Every industrial strength database I've used, prior to PG, does this. Some better than others, of course; as Baron mentions. How much effort is needed, I don't know, as I don't write database engines, only use them.Robert Younghttps://www.blogger.com/profile/09056808374481236610noreply@blogger.comtag:blogger.com,1999:blog-20038672.post-26321964729395170162010-11-11T08:46:24.420-05:002010-11-11T08:46:24.420-05:00The theoretical limit is 2 billion for vacuum free...The theoretical limit is 2 billion for vacuum freeze, but we default to 200 million so we can trim pg_clog, so our limiting factor is not the xid size but the pg_clog size.Bruce Momjianhttps://www.blogger.com/profile/07875088787463864011noreply@blogger.comtag:blogger.com,1999:blog-20038672.post-71083868030283700242010-11-11T08:41:00.369-05:002010-11-11T08:41:00.369-05:00From the video of the talk, I got the impression R...From the video of the talk, I got the impression Rob's comments about Postgres were driven mostly by reading mailing list postings or other articles rather than actual hands-on work. Hence, your being baffled by his "statistics issues."Joehttps://www.blogger.com/profile/18006904787527262772noreply@blogger.comtag:blogger.com,1999:blog-20038672.post-35937767384417062732010-11-11T07:31:19.838-05:002010-11-11T07:31:19.838-05:00I disagree that the query cache is something MySQL...I disagree that the query cache is something MySQL does well. I wouldn't try to add this to Postgres. It causes lots of performance problems.Baronhttps://www.blogger.com/profile/01621441847303652718noreply@blogger.comtag:blogger.com,1999:blog-20038672.post-17325323252388379892010-11-11T00:42:04.414-05:002010-11-11T00:42:04.414-05:00One thing I got a kick out of his talk were the ma...One thing I got a kick out of his talk were the many times he talked about various mysql points, then made the comment about it has issues, followed by "it just works". I lost count how many times he dropped those two phrases together. Of course, my take is simple: "it has issues" != "it just works".Anonymousnoreply@blogger.com