Hello, I noticed several times that performances appeared to be varying a lot over remote connections. May I ask for a python script to benchmark the connection and the server overall reactivity ? It would play a typical scenario corresponding to typical users, and would provide objective performance ratings in "Tiny-ops", characterizing critical sequences : connection, database navigation, large search requests, data updates... Because the script performs always the same way we would be able to mesure improvements and compare results. It would also be useful to simulate many simultaneous users if it loops with different test profiles and if we can start it many times on several clients.
Guest wrote: > > wether XML-RPC is at its end of speed, > what alternatives you evaluated or thought about > wether "2 - 8 times" can really be regarded an improvement. Will the > connection be fast or less slow? > what could be done with 2000 or 5000 EUR to improve speed? I agree with you, nevertheless, my bet is that we'll know whether it improve the global picture when we'll have it for playing in real environments. French typical ADSL connections ping in 50ms, which is 50-200 times slower than lan. It downloads 8 to 10 mbps, which is quite good. Bandwith is usually not significant, as far as I've seen. As a comparison, I experimented sometimes that a Tiny form (partner or product) is presented 10-20 times slower than a web page, at the same time and from the same server. And at a later time, it's quite ok to work with... Of course for large messages, benefit is obvious, but I m not sure how often such messages are used. Therefore I hope that we will save IP packets when message size is reduced, because some messages may be cut in 2 or 3 packets today, while they would fit in a single packet with this compression. If response time average is 20% better, it's worth doing it. Less than 10% would be disappointing. I'm not convinced that all other potential improvement areas have been visited to optimize protocol, but let 's start by this end. ------------------------ Dominique www.SISalp.org _______________________________________________ Tinyerp-users mailing list http://tiny.be/mailman/listinfo/tinyerp-users
