Adrian Chadd disse na ultima mensagem: > On Mon, Jun 18, 2007, Michel Santos wrote: > >> yes I tried it. Even if the aufs threads are running on all CPUs and it >> gives a better performance in comparism to ufs, still, diskd runs >> faster, >> specially under load and lots of connections. Response time of squid >> with >> diskd under load is much better here. > > Then someone's going to have to do some digging into fixing the issue. > To be honest, someone should do some digging into the last ten years of > HTTP cache filesystem work and just write something to replace COSS > that isn't so temperamental and difficult to configure; I want to do it > but I just haven't any time anymore. > > (and then contribute it back to Squid, unlike 100% of the papers I've > read on this subject which used Squid in their development and testing > phases.) so if I had enough money I would even spend on this project to help the "time issue" out :) ... so if there is a developer who does not pretend to charge as ibm or cisco consulting can contact me to fiddle something out anyway, I have something for you Adrian ... the NASA developed in order to sincronize earth-marth teams a clock for the MARS Rover expedidition where the day is 39 minutes shorter (on Mars) so you take one of this clocks running Mars-Time and you get almost an hour extra each day! And in exchange for this particular benefit you make COSS starting faster for me ok? Michel ... **************************************************** Datacenter Matik http://datacenter.matik.com.br E-Mail e Data Hosting Service para Profissionais. ****************************************************
