Return-Path: Mailing-List: contact jmeter-user-help@jakarta.apache.org; run by ezmlm Delivered-To: mailing list jmeter-user@jakarta.apache.org Received: (qmail 37063 invoked from network); 25 Sep 2003 12:41:10 -0000 Received: from unknown (HELO dent.mvlan.net) (192.197.166.10) by daedalus.apache.org with SMTP; 25 Sep 2003 12:41:10 -0000 Received: from zaphod.mvlan.net (zaphod.lan11.localdomain [10.1.1.160]) by dent.mvlan.net (Postfix) with ESMTP id 79426E1B3CF for ; Thu, 25 Sep 2003 08:41:11 -0400 (EDT) Received: by zaphod.mvlan.net (Postfix, from userid 500) id 838725022B3; Thu, 25 Sep 2003 08:41:09 -0400 (EDT) Date: Thu, 25 Sep 2003 08:41:09 -0400 From: Jean-Sebastien Morisset To: JMeter Users List Subject: Re: ab vs jmeter Message-ID: <20030925124109.GA12729@zaphod.mvlan.net> References: <20030923205821.GA20648@zaphod.mvlan.net> <20030924151853.50537.qmail@web41712.mail.yahoo.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20030924151853.50537.qmail@web41712.mail.yahoo.com> User-Agent: Mutt/1.4i X-Priority: 3 (Normal) X-PGP-Key-ID: 0xC61A8A78 X-PGP-Key-URL: http://jsmoriss.mvlan.net:8080/pgp-public-keys.txt X-PGP-Key-Fingerprint: 9367 234C 4092 0E7C 09DD 55FA 9C6C 8839 C61A 8A78 X-OS: Linux zaphod 2.4.19-xfs #6 Wed Sep 17 16:45:50 EDT 2003 i686 unknown X-Uptime: 7:48am up 2 min, 1 user, load average: 1.62, 0.61, 0.21 X-Seti: Name=jsm-mv #WU=5792 CPU=2951 days (Best Scores: Spike=1.280096 Gaussian=0.417315) X-ICQ: 47027114 X-Spam-Rating: daedalus.apache.org 1.6.2 0/1000/N On Wed, Sep 24, 2003 at 08:18:53AM -0700, Frank Rizzo wrote: > you should be able to run at least 20 threads on a pentium class machine. also, if you click on the test plan node in the tree, make sure functional test mode is unchecked. > [snip!] > > So the answer is that jmeter, under normal circumstances, should be able to perform better than what you've reported. But without more information, I cannot speculate why you are seeing such poor performance. > Thanks for your reply Frank. With 2-3 threads on my smallest machine, jmeter-server maxed out the CPU, while ApacheBench could run with 1000 concurrent requests. Admitedly, this was a very small machine and is not representative of the hardware I'll be using to run the actual tests. It _did_ give me a good comparison between AB and JMeter. At first it appeared as though the stress test would be simple GETs, but I just learned that for each "session" (1 get of this page, 5 gets of that one, and 10 of another) I must supply a unique 32 bit alpha-numeric parameter in each request. ApacheBench cannot do this, so I'll have to use JMeter. I've found a couple of decent size servers I can use, so it shouldn't be too bad. My problem will be generating that 32 bit string. The "Counter" pre-processor certainly won't be enough. Do you have any suggestions? :) Thanks! js. -- Jean-Sebastien Morisset, Sr. UNIX Administrator Personal Home Page JS & Melanie's Homebrewery Underwater and Travel Photographs