On Friday 08 December 2006 17:48, Robert Nelson wrote: > No I haven't incorporated the gigaslam test yet.
Yes, if you do, be careful to not leave any copies of gigaslam.gif hanging around in the regression directory -- my first few cuts did that and I thought my kernel was going south because I couldn't do *anything* while the regression scripts were running. It turns out that reading a 1GB sparse file with only two real blocks and writing all those zeros with Bacula makes your desktop response time catastrophic. The cleanup script now takes a lot more care to delete all gigaslam.gif files. > > This is just the standard non-root file tests. OK. > > I'll take a look at those two pragmas and see if that helps. > > More than 10 times slower is what I'm seeing. Yes, that is pretty disasterous. If the 10 times is the overall speed, then the database slowdown is even worse because I estimate that the database part of the regression tests is less than 30% of the total time -- I have no hard data to back that up though. > > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of Kern > Sibbald > Sent: Friday, December 08, 2006 8:43 AM > To: bacula-users@lists.sourceforge.net > Cc: Robert Nelson; 'bacula-devel' > Subject: Re: [Bacula-devel] [Bacula-users] What a difference a database > makes > > On Friday 08 December 2006 16:21, Robert Nelson wrote: > > I've been working on the regression tests for the imminent release of > 1.40. > > Initially I was using SQLite3 since it has zero maintenance and zero setup > > and the tests aren't all that strenuous (at least as far as the database > is > > concerned). > > > > > > > > However, I found a stress run was taking 12 hours on a dual proc machine > > with 2 GB of memory. That seemed rather outrageous so I tried with MySQL > > and it dropped to an hour and a half. > > Is that the gigaslam test? > > > > > > > > > Is SQLite3 really this bad? I can't imagine so or it wouldn't survive. > In > > that case, there must be something that we are doing or not doing in > Bacula > > or some problem in our build options that is killing the performance. > > Well, in some really rough measurements that I did, SQLite version 2 was > roughly the same speed as MySQL for the regression scripts. There have been > some reports that SQLite might suffer with really big databases compared to > MySQL. > > My equally rough tests with SQLite3 showed it was 4-10 (possibly more) times > > slower than SQLite version 2. I've never taken the time to try to find out > why, so I have never mentioned it to the author. For that reason, I always > use SQLite version 2 for my zero maintenance tests -- I also run the > regression scripts prior to major releases on MySQL and PostgreSQL, and my > "production" database is MySQL. > > > > > > > > > Is anyone out there using SQLite3? Does anyone have any suggestions on > how > > we could change the default build defines or database creation scripts to > > improve the performance? > > > > > > > > If there are any suggestions I'd appreciate getting them quickly so that > we > > can incorporate them into 1.40. > > When SQLite3 was first released, I looked into this and seems to me that he > handled synchronization differently. The two pragmas that I saw (long ago) > > that made a difference were: > > PRAGMA default_synchronous > PRAGMA default_cache_size > > > ------------------------------------------------------------------------- > Take Surveys. Earn Cash. Influence the Future of IT > Join SourceForge.net's Techsay panel and you'll get the chance to share your > opinions on IT & business topics through brief surveys - and earn cash > http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV > _______________________________________________ > Bacula-devel mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/bacula-devel > > > > > ------------------------------------------------------------------------- > Take Surveys. Earn Cash. Influence the Future of IT > Join SourceForge.net's Techsay panel and you'll get the chance to share your > opinions on IT & business topics through brief surveys - and earn cash > http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV > _______________________________________________ > Bacula-users mailing list > Bacula-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/bacula-users > ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys - and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
