Gerhard | Morning all | 00:18 |
---|---|---|
*** Gerhard has quit IRC | 00:42 | |
*** Alex|off is now known as Alex| | 01:48 | |
*** cthibert has joined #ingres | 03:31 | |
*** atrofast has joined #ingres | 04:52 | |
*** Gerhard has joined #ingres | 04:52 | |
*** Gerhard has quit IRC | 06:13 | |
*** rossand has joined #ingres | 06:15 | |
*** ChanServ sets mode: +o rossand | 06:15 | |
*** DarylM has joined #ingres | 06:23 | |
pboro | Ingres is really slow when doing a lot of operations using JDBC... I would expect it to be I/O intensive, not CPU intensive :( | 06:48 |
pboro | Is there an easy way to profile Ingres? | 06:48 |
atrofast | pboro: Have you tried OProfile? (http://oprofile.sourceforge.net/) | 06:52 |
pboro | nope... have you tried it with Ingres? | 06:52 |
pboro | Hmm it prolly requires that I have a -g version of Ingres | 06:55 |
atrofast | Only very shortly yeah... but as you said I had compiled it with -g | 07:01 |
atrofast | Not sure if it'll work without that since I haven't tried :) | 07:02 |
pboro | yeah, was it easy to use? | 07:02 |
pboro | I guess I'll have to make -g compiled version of latest 9.2 and profile it a bit, there seems to be something fishy in the performance | 07:02 |
pboro | ahh... I guess these instructions could be useful: http://wiki.postgresql.org/wiki/Profiling_with_OProfile | 07:03 |
pboro | hmmh... I guess it's no use to profile 9.2 anyway, instead I should use at least 9.3 | 07:09 |
pboro | it would be nice if the community versions were also provided as debug enabled (-g) versions :) | 07:10 |
*** cthibert has left #ingres | 07:11 | |
atrofast | I love debug enabled versions :D | 07:12 |
atrofast | I wish I could help you more, I only did some very quick tests with oprofile but it seemed fairly easy to use, it's a daemon that runs in the background collecting information about current processes and so on | 07:13 |
pboro | ahh, Ingres 9.3 for Windows 32-bit is now available at ESD | 07:13 |
pboro | atrofast, yeah, thanks :) | 07:13 |
*** grantc has joined #ingres | 07:17 | |
*** ChanServ sets mode: +o grantc | 07:17 | |
grantc | hi pboro - you should be able to get the GA 9.3 from esd | 07:18 |
pboro | hm, for 9.3... "Support for Multiple Data Access Servers", but why? :D does it provide more performance or what? | 07:18 |
pboro | yeah, currently testing it :) | 07:18 |
pboro | grantc, do you have experience on profiling Ingres? | 07:21 |
grantc | not in the last 5+ years | 07:22 |
grantc | i used to do some performance tuning stuff but i have forgotten most of it now | 07:22 |
pboro | I have noticed that iidbms uses pretty much CPU when using JDBC, even when doing things that I would expect to be I/O intensive... so I figured, I could do some profiling to see what's going on | 07:23 |
pboro | I guess I'll compile debug enabled (-g) version of latest 9.3 community and try running oprofiler or similar to get some numbers | 07:23 |
grantc | you might be better off asking withdefault the next time he surfaces here | 07:23 |
pboro | thanks for the tip | 07:24 |
grantc | he might not know either but he has been involved in some projects with Windows and JDBC | 07:24 |
bonsaikitten | btw, the build system is quite funny. parallel building is definitely not (yet) supported | 07:24 |
bonsaikitten | needs more love :) | 07:24 |
pboro | grantc, I've been trying my project's software on PostgreSQL (to get some benchmark comparison) and it performs a lot better, so I guess there's something that could be done better in Ingres :) | 07:26 |
pboro | and I have journaling etc. turned of from Ingres so it should not matter | 07:26 |
pboro | and with "a lot better" I mean differences like >200% in speed... | 07:28 |
pboro | it could, of course, be a configuration problem too :) | 07:28 |
grantc | hmm - I'll mention it to my boss as it's her team that looks after DAS | 07:28 |
grantc | i have a call with her in the next hour | 07:28 |
pboro | well no need really, until I get some numbers, it could be anything really :) | 07:29 |
pboro | my software only works through JDBC so I can' t compare with other connectivity means | 07:29 |
pboro | but thanks | 07:29 |
grantc | ok - i assume that you have cursor=readonly set, either as a property or attribute (if not then the newer drivers _should_ have it set if not defined) | 07:30 |
pboro | I'm using iijdbc 3.4.8, it should have it by default | 07:31 |
pboro | +set | 07:31 |
grantc | fair enough. I'll mention it to my boss anyway - if you do get some numbers we can look at you can you mail them over? | 07:32 |
pboro | sure, I'm more than interested in getting Ingres better in any aspect :) | 07:32 |
grantc | cool :) | 07:33 |
pboro | any other profilers to recommend than oprofiler? it was recommended by atrofast... what tool does Ingres Co. use for profiling? | 07:33 |
grantc | not sure - i know it's not cheap, whatever we use | 07:34 |
pboro | ahh, some commercial product then :) | 07:35 |
grantc | yup - to be honest i don't do much code profiling on my stuff but I seem to remember intel vtune being mentioned (having jogged my brain cell) | 07:37 |
grantc | and quantify/purify | 07:37 |
pboro | hehe, ok | 07:37 |
*** Alex| is now known as Alex|off | 08:02 | |
*** Deyan has joined #ingres | 08:59 | |
Deyan | hello | 08:59 |
bonsaikitten | g'day | 09:01 |
*** monda07 has joined #ingres | 09:07 | |
grantc | pboro: do you have cache_dynamic enabled for your testing? | 09:08 |
pboro | grantc, no, last time I used it in production it caused some weird crashes | 09:10 |
pboro | has it been fixed already? | 09:10 |
*** FrankW has joined #ingres | 09:10 | |
grantc | there may be some edge conditions in 9.3 but it's a lot better | 09:10 |
grantc | specifically LOB deadlocks | 09:10 |
pboro | yeah, those bite me last time... | 09:11 |
grantc | :( | 09:11 |
grantc | if you can try with it enabled for your 9.3 testing I would be interested to see if it improves things | 09:11 |
pboro | I'll try it tomorrow, I'm at home already :) thanks for the tip | 09:12 |
*** cthibert has joined #ingres | 09:14 | |
*** DarylM has quit IRC | 09:16 | |
*** monda07 is now known as DarylM | 09:16 | |
grantc | ta | 09:22 |
*** grantc has quit IRC | 09:28 | |
*** cthibert has left #ingres | 09:46 | |
*** cthibert has joined #ingres | 09:54 | |
*** rilson has joined #ingres | 10:02 | |
*** Alex|off is now known as Alex| | 10:13 | |
*** Deyan has quit IRC | 10:17 | |
*** Alex| is now known as Alex|off | 10:27 | |
*** monda07 has joined #ingres | 12:48 | |
*** cthibert has left #ingres | 12:57 | |
*** DarylM has quit IRC | 12:57 | |
*** monda07 is now known as DarylM | 12:57 | |
*** monda07 has joined #ingres | 14:00 | |
*** DarylM has quit IRC | 14:09 | |
*** monda07 is now known as DarylM | 14:09 | |
*** rilson has quit IRC | 15:17 | |
*** atrofast1 has joined #ingres | 15:32 | |
*** atrofast1 has left #ingres | 15:32 | |
*** atrofast has quit IRC | 15:35 | |
*** DarylM has quit IRC | 15:42 | |
*** rossand has quit IRC | 17:19 | |
*** rossand has joined #ingres | 17:45 | |
*** ChanServ sets mode: +o rossand | 17:45 | |
*** rossand has quit IRC | 17:45 | |
*** rossand has joined #ingres | 18:04 | |
*** ChanServ sets mode: +o rossand | 18:04 | |
*** rossand has quit IRC | 18:43 | |
*** DarylM has joined #ingres | 21:48 | |
*** Alex|off is now known as Alex| | 22:36 | |
*** zxiiro has quit IRC | 23:00 | |
*** zxiiro has joined #ingres | 23:01 | |
*** ChanServ sets mode: +v zxiiro | 23:01 | |
*** Alex| is now known as Alex|off | 23:09 | |
*** Alex|off is now known as Alex| | 23:24 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!