> I have found that the main bottleneck is not in code interpretation (the pure > computation speed), but: > 1) network (DB <-> AS <-> Client); > 2) not optimized SQL queries; > 3) not optimized DB architecture (played a little bit with indexes and got > huge performance boost); > 4) fine tuning PGSQL maintenanace (reindex, vacuum); > 5) implement some ORM functionality as PGSQL triggers and built in functions > (have not tried this by myself, but as i have read PGSQL allows writing > extensions in python as well in other languages)
Yes, I think the database and the network are the two bottlenecks. For the network, the new version of the web client will solve the problem. (to be released in about 2 months) I am interested by the optimization you made on the database so that we can include it in the core of Open ERP. Was in on version 4.2 or version 5.0 ? Because I think indexes are good in version 5 ? We are currently fine tuning for a customer that works on Open ERP with 5.000.000 of customers, all improvements we will ahve to do will be reintegrated in the core. For points 5, I have a lot of doubts about these improvements. I also think python is not a problem, on a not too small database, your cpu becomes: 90% postgresql, 10% python. We have found a 'speed bug' introduced in v5 for the reporting, we will release a new version soon. We already have 60x faster for the same reports (printing 500 sales orders) -------------------- m2f -------------------- -- http://www.openobject.com/forum/viewtopic.php?p=34375#34375 -------------------- m2f -------------------- _______________________________________________ Tinyerp-users mailing list http://tiny.be/mailman/listinfo/tinyerp-users
