Hi,
As some of you have already seen, I'm a new commiter and I haven't been idle theese days. With my systems experience hat on and wikipedia loving heart inside I started porting Mediawiki software to PostgreSQL.
Well, the starting (and telling out) point is already kind of working Mediawiki 1.3 installation on http://ph3.defau.lt/. Most of it is commited already into project CVS, but some broken (and already fixed) parts are awaiting 1.3 official release, so I could break things again (oops, it isn't my mission :), as I've already did, and thanks to Tim and Brion for quick fixes. Some critical patches are in http://dammit.lt/pgwiki/20040613-diff.txt.
For people, not familiar with PG I can do some easy marketing :) 1. It's fast. Well, if you've heard stories about slow elephant (represented by logo:), it has really changed. Modern PG versions are fast. Especially under high concurrent loads. It also greatly performs in joins - may use more than one index, has intelligent query optimiser, etc. 2. It has features. Stable ones :) Triggers, stored procedures, subselects, ... 3. It is stable enough.
Anyway. Right now I started a testbox, on which I do some performance testing, I started posting the results of tests on http://ph3.defau.lt/index.php/Performance_testing. I'd be glad to find out what queries you would like to observe on this testbox, so please leave comments everywhere :)
The production-quality PG port is not too far away. :)
Cheers, Domas
P.S. I don't use MW for any other than wikipedia's purposes, so please note, that without wikipedia's feedback I don't be able to fix things :)
Domas Mituzas wrote:
Hi,
As some of you have already seen, I'm a new commiter and I haven't been idle theese days. With my systems experience hat on and wikipedia loving heart inside I started porting Mediawiki software to PostgreSQL.
Well, the starting (and telling out) point is already kind of working Mediawiki 1.3 installation on http://ph3.defau.lt/. Most of it is commited already into project CVS, but some broken (and already fixed) parts are awaiting 1.3 official release, so I could break things again (oops, it isn't my mission :), as I've already did, and thanks to Tim and Brion for quick fixes. Some critical patches are in http://dammit.lt/pgwiki/20040613-diff.txt.
For people, not familiar with PG I can do some easy marketing :)
- It's fast. Well, if you've heard stories about slow elephant
(represented by logo:), it has really changed. Modern PG versions are fast. Especially under high concurrent loads. It also greatly performs in joins - may use more than one index, has intelligent query optimiser, etc. 2. It has features. Stable ones :) Triggers, stored procedures, subselects, ... 3. It is stable enough.
Anyway. Right now I started a testbox, on which I do some performance testing, I started posting the results of tests on http://ph3.defau.lt/index.php/Performance_testing. I'd be glad to find out what queries you would like to observe on this testbox, so please leave comments everywhere :)
The production-quality PG port is not too far away. :)
Cheers, Domas
Domas,
This is excellent news! PG may well provide the extra scalability that Wikipedia will need soon.
Regards,
-- Neil
wikitech-l@lists.wikimedia.org