tag:blogger.com,1999:blog-9144505959002328789.post3976018464542483776..comments2024-03-28T11:39:50.622+01:00Comments on Karlsson on databases and stuff: New Storage Engine Kids on LinuxKarlssonhttp://www.blogger.com/profile/04874338187076980133noreply@blogger.comBlogger5125tag:blogger.com,1999:blog-9144505959002328789.post-18176863416088571842010-01-09T10:35:45.056+01:002010-01-09T10:35:45.056+01:00Swany: I agree, to an extent. But the idea was to ...Swany: I agree, to an extent. But the idea was to compensate for this, to an extent, by testing something that is as little effected by tuning as possible, and for that I went for a mass-INSERT test.<br /><br />Having said that, I still don't consider this to be conclusive at all, but in my mind, no benchmark is conclusive, and performance, per se, is not that interesting to me, once we get past a certain level. You can always add more machines, do your operations differently, shard your data etc. So I just wanted to test something silly out of the box.<br /><br />Also, in this case, I found no good generic tuning samples for Maria or PBXT, no sample my.cnf that came with MariaDB contained any Maria or PBXT parameters at all. You could draw any kind of conclusion from that of course.<br /><br />I'd be happy to do an optimal tuning test, but that will require a fair bit more effort, so I don't think I have the bandwidth for that. In the case of Maria though, I'll make an exception as it was so far behind the other guys.<br /><br />Anyone who want to send me some optimal mass-INSERT tunings for the other engines are also welcome, and I'll test again, and do a third blogpost with the results next week.<br /><br />/KarlssonKarlssonhttps://www.blogger.com/profile/04874338187076980133noreply@blogger.comtag:blogger.com,1999:blog-9144505959002328789.post-22358684169617020772010-01-09T10:09:31.853+01:002010-01-09T10:09:31.853+01:00Benchmarking with absolutely no tuning is an exerc...Benchmarking with absolutely no tuning is an exercise in futility.<br /><br />You can't expect an engine to have excellent settings for any arbitrary benchmark.Justin Swanharthttps://www.blogger.com/profile/08193089637089861226noreply@blogger.comtag:blogger.com,1999:blog-9144505959002328789.post-1666767487817705482010-01-08T04:11:11.991+01:002010-01-08T04:11:11.991+01:00Benchmarks without context don't do much to ad...Benchmarks without context don't do much to advance the discussion. I can write many more rows per second to the /dev/null storage engine. I will guess that by 'sustained' you mean 'sustained' when there are no indexes.Mark Callaghanhttps://www.blogger.com/profile/09590445221922043181noreply@blogger.comtag:blogger.com,1999:blog-9144505959002328789.post-25791137618807776302010-01-08T02:55:45.518+01:002010-01-08T02:55:45.518+01:00It all depends on how you test, and different engi...It all depends on how you test, and different engines are suitable for different purposes but they also need to be handled differently for that purpose by the app.<br /><br />For instance, I can insert 340,000 rows per second, sustainable, into a MyISAM table from a single thread on a basic 64bit Linux. Unmodified server code, just appropriate tuning and SQL approach.<br /><br />Just to give a bit of perspective on your 4,000 operations with MyISAM ;-)arjenhttps://www.blogger.com/profile/01398380484205552972noreply@blogger.comtag:blogger.com,1999:blog-9144505959002328789.post-26597664146629257052010-01-07T15:55:23.470+01:002010-01-07T15:55:23.470+01:00I too hope for a better MyISAM. But Maria is tryin...I too hope for a better MyISAM. But Maria is trying to do much more and be an alternative to InnoDB. It is much easier to be a better MyISAM than a great OLTP engine. So we must wait until someone decides to fix some of the problems in MyISAM and try to ignore all of the money that MySQL invested in Maria and Falcon.<br /><br />Given the current state of MyISAM I am amused that they added support for hot backup to it. If you need hot backup, why would you use MyISAM?Mark Callaghanhttps://www.blogger.com/profile/09590445221922043181noreply@blogger.com