Lee Chin wrote: > > But why would they be measuring IIS and tweaking web servers like Apache if > all they are doing is measuring connections acknowledged? Thats an OS > thing! Beats me.... Here's part of the webbench documentation describing the requests/second test: "Requests per seconds are a basic measure of the client/server interaction and give you a rough idea of how many hits per day a server can handle. Keep in mind, though, that WebBench's requests per seconds score is generally higher than the typical load serviced by most Web servers because WebBench uses stress tests to bombard the server with requests as fast as the clients can issue them." Hmmm.... testing the server's ability to cope with a DOS attack? > > This is whacky Could be it's the only thing IIS does really well. -jdr- > ------Original Message------ > From: Jim Reimer <jim.reimer@asec.delphiauto.com> > To: Lee Chin <leechin@mail.com> > Sent: April 5, 2001 12:41:54 AM GMT > Subject: Re: How do we interpret this ZDNET benchmark? > > Looks to me like it's the number of requests acknowledged, not served. > > What I don't understand is how a single real-world client could possibly > generate more than 100,000 requests per second, unless it's engaged in a > DOS attack or something similar. (???) > > Maybe I'm wrong, but it appears to be somewhat contrived. > > -jdr- > > Lee Chin wrote: > > > > If you look at the graph at the bottom of the page, they say that their > > windows 2000 box can perform over 1 million requests per second with IIS! > > > > http://www.zdnet.co.uk/pcmag/ne/2000/11/01.html > > > > Now, I may be naive in my math (and if so I need to know why) but on a > > machine that has a 100 Mbps card, if I serve 1 million pages in one > second, > > that means my page size (web page Im serving) is > > > > 100000000 (ie. 100 Mb) / 1 Million = ~ 13 bytes > > > > So if they are saying they did 1 million requests per second on a 100 Mbps > > connection, then that means their web pages were 13 bytes long.... > > > > Are they nuts or am I interpreting the graph wrong? > > > > Thanks > > Lee > - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org
