> Thanks for your explanations Roberto! Just a few follow-up questions. > > t on 32bit system you have >> at >> max a couple of gigs of memory for process) and the not-perfect >> isolation >> of python interprerers. For me, 20 apps are too much for a single >> process, >> i suggest you to avoid this approach > > Understood. Considering these limitations, in which situation would > you recommend the usage of the "multiple interpreter in the same > process" setup?
I always try to avoid having (on 32bit systems) processes with more than 512M of memory, if you host tiny apps it should be enough, but honestly this is only a my 'paranoia' :) I have instance that hosts upto 8 apps but i have never tried with more. > >> >> http://projects.unbit.it/uwsgi/wiki/CustomRouting >> > > Ok. I am reading the wiki page and will give a thought about it. What > is the expected overhead of the "indirection"? > > >From my benchmark usign the 'virtualhosting' example in the wiki you have an impact of the 2% (no real differences between tcp and unix sockets). If someone propose a standard for virtualhosting-proxy, the need of having to write python code for routing can be removed, gaining a lot of more power. In the mean time i am extending the uwsgi api to support multiple requests in the same connection, in this way the custom-router can connect to instances on startup an then reuse it for all the requests (removing the overhead of connect()/accept()) -- Roberto De Ioris http://unbit.it _______________________________________________ uWSGI mailing list [email protected] http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
