your way, right away
|
saidone.org -
|
|
Main menu
Home
About Projects
AlfMarkSwamp
Links
ush.itaghers.org saidone@ush gameknot virtualmagister Misc
Unless otherwise specified, all contents of this site are licensed under a Creative Commons Attribution-ShareAlike 3.0 Italy License.
 |
AlfMark - measuring database performance - posted by saidone on Thu, 01 Sep 2011 08:59:40 GMT
As a test for current (crippled at the moment, I know) version of AlfMark I wanted to try to squeeze the most of performance from the usual test machine.I tried some sort of optimizations on MySQL: disabling checksums, double writes and log flush on transaction commit does not help, neither on the tests with more concurrency (and note that we are loosing ACID compliance!). Also increasing the various buffer and cache sizes, that usually help a lot, had not significant impact on benchmark results, and probably a new set of tests will needed to put heavy stress on database. No speed gain also with slight tuning on table indexes. After that I decided to switch database, from MySQL to PostgreSQL (8.4.7, as installed from CentOS 6 repositories) and relaunched the benchmark. And, surprise!, score increased by 15%: I've almost never used or deployed PostgreSQL (apart from an old project that involved a porting of large portions of PL/SQL from Oracle) but this boost came unexpectedly, having heard tens of MySQL advocates telling me that it was at the top when speaking of raw speed performance. At least for Alfresco this is no more true, don't matter if this came from misconfigurations of different tuning or something else, but, with out of the box configuration, that's it. Tweet No comments yet. Post a new comment back to home |
[[[[[[[ served by kugelmass ]]]]]]]
If I had only known, I would have been a locksmith. - Albert Einstein
|
|