Date: Sat, 20 May 2006 18:25:14 +0100 (BST) From: Robert Watson <rwatson@FreeBSD.org> To: Ivan Voras <ivoras@fer.hr> Cc: performance@freebsd.org, Michael Vince <mv@thebeastie.org> Subject: Re: (Another) simple benchmark Message-ID: <20060520182238.R8068@fledge.watson.org> In-Reply-To: <446DE927.2060909@fer.hr> References: <446CCE1C.1050200@fer.hr> <446CD873.9080903@stevehodgson.co.uk> <446CE6CE.50009@fer.hr> <446D8994.3070709@thebeastie.org> <446D9DEE.4050300@fer.hr> <446DE1F2.4020602@thebeastie.org> <446DE927.2060909@fer.hr>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, 19 May 2006, Ivan Voras wrote: > Michael Vince wrote: > >> What I am trying to say here is you are expecting good performance out of >> things like CGI/PHP and prefork, > > Ok, did anybody read my initial post? > > I'm NOT setting up a production machine. I'm NOT using PHP - it was > mentioned as a reason threaded apache is not widely used. I've run "ab -n > 100000 -c 100 http://localhost/" where "http://localhost/" references a > small static HTML file, served by apache 2.0.x. PHP was not even included in > both apache setups. I've run this on a 8-cpu Xeon beast (ok, not really - 4 > cpus, tried with hyperthreading on and off) and got terrible performance. > This performance is objectively low even by itself, without any comparison > with other operating systems (such as linux). > > What I *am* doing now is looking for someone who has a 4 CPU or bigger > machine idle on which he/she can replicate this simple benchmark (it really > IS simple - you need apache20 port in default configuration - everything's > included) and confirm or contradict my results. I won't tell exactly what my > results are because: a) to encourage fairness and b) because they are so > ridiculously low that if I'm wrong I don't want it to end up in mailing list > archives for posterity :) What I'd like to see someone do, perhaps you, is the following: - Configure one system as the apache benchmark client system. Optimize the heck out of it until you can generate HTTP queries just as fast as you possibly can. If it's Linux, FreeBSD 4, FreeBSD 6, Windows, Solaris, whatever, it doesn't matter, as long as you can generate them really fast. - Now configure a single system or set of identical systems as web servers using the configuration permutations of choice. - Test the various server configurations with the single client configuration. I'm set up to easily benchmark FreeBSD in our netperf cluster, but I'm not set up to easily do comparisons with Linux, since we don't have a pluggable netboot Linux to install in that environment. If someone wants to provide some pointers doing that, I'm happy to explore doing it, but ideally it would be something like "tftp export file <a> and NFS tree <b>, boot". This is the model we use for FreeBSD in the netperf cluster, and it works very well :-). Robert N M Watson
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060520182238.R8068>