Return-Path: Delivered-To: new-httpd-archive@hyperreal.org Received: (qmail 407 invoked by uid 6000); 15 Jun 1998 20:29:16 -0000 Received: (qmail 400 invoked from network); 15 Jun 1998 20:29:14 -0000 Received: from ns2.gamespot.com (HELO mail.gamespot.com) (206.169.18.3) by taz.hyperreal.org with SMTP; 15 Jun 1998 20:29:14 -0000 Received: from localhost (ian@localhost) by mail.gamespot.com (8.9.0/8.9.0) with SMTP id NAA29096 for ; Mon, 15 Jun 1998 13:29:07 -0700 (PDT) Date: Mon, 15 Jun 1998 13:29:07 -0700 (PDT) From: Ian Kallen To: new-httpd@apache.org Subject: Re: Apache In The News: Latest PC Week Article In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: new-httpd-owner@apache.org Precedence: bulk Reply-To: new-httpd@apache.org Temporarily delurking here... I did some benchmarking recently and was surprised at how high IIS and Enterprise Server came out. Just like MS and Sun torque their JVM's to do well on the CaffeineMark benchmarks, I suspect they similarly optimize to handle the extremely artificial environment of benchmarks. For the article I wrote up, I emphasized specifically that benchmarks measure one thing and it's one thing that is not very important to me: how well the server handles benchmark loads. If folks here are really worried about this then developing algorithms that optimize handling large numbers of GETs from a proportionately small number of clients on low latency networks should be a priority. Otherwise, take it for what it is: measurements that have little bearing on real world performance. It's not that I don't think the benchmark authors are sincere, I just don't think they really understand the number of crazy factors that effect real world performance: network latency, buggy browsers and goofy tcp/ip implementations are not simulated. -Ian On Mon, 15 Jun 1998, Rasmus Lerdorf wrote: :> Your 500 requets/sec qare a _lot_ easier than theirs probably. :> :> This is over a large dataset (at least a few hundred megs) of varying :> sizes. : :Could be. They don't expand on that aspect of it. If IIS was twice as :fast as Apache on similar hardware, then I don't think I would be getting :all these NT/IIS/ASP to Unix/Apache/PHP converts showing up on the PHP3 :mailing list. Most of them cite performance and stability as their reason :for switching. -- Where there's no emotion, there's no motive for violence. -- Spock, "Dagger of the Mind", stardate 2715.1