Date: Sun, 26 Apr 1998 14:22:06 -0600 (MDT) From: Marc Slemko <marcs@znep.com> To: Don Wilde <dwilde1@ibm.net> Cc: freebsd-chat@FreeBSD.ORG Subject: Re: A name, please! Message-ID: <Pine.BSF.3.95.980426140131.12990B-100000@alive.znep.com> In-Reply-To: <35437894.CFD08DE6@ibm.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, 26 Apr 1998, Don Wilde wrote: > How about "The Great Free World Web Speed Challenge"? > > It's not a shoot-out yet, that will happen when we get everybody > together, and since there will be complaints that we use unfair > hardware, we will get our wish that the configurations will be more in > our normal working grounds. > > We, the "free software" world are challenging the corporate world with > our web speed. It's a speed test, and we're throwing the gauntlet right > square in the face of Microsoft, Novell, IBM, HP, Sun, and everybody > else. And you will get it thrown right back in your face and have your ass kicked badly if you approach it this way. I'm really trying not to be negative about this, but you are a bit heavy on enthusiasm and a bit light on substance. 1. You want to use the most expensive hardware possible, solid state disks, etc. which you envision costing up to $100k for the box. At that level, you are competing against very big toys and I don't think you fully understand what you are up against or the time and money devoted by vendors to get the results that you see published and that you think you can beat. 2. You want to make a media circus out of something that you have no experience with, no test runs, no time for the way benchmarking is really done. It isn't a thing you can make fancy and glamorous that easy; the fact is that 99% of the work in benchmarking is ugly slogging that takes time. During this time there is nothing for people to see, no wild cheering, etc. This is the perfect formula for a big flop that ends up making a joke out of FreeBSD, Apache (or whatever web server you use), and gives the media a good laugh. 3. You are presuming you can get all this hardware together (remember you need some pretty big client machines as well to pull off the benchmark) without trouble, put it all together and make it work in the span of a couple of days, etc. If you instead tried to find and work a deal with a company with a lab that already has most of the setup necessary, you would probably have more luck. If you went for a half dozen average (aka. pretty cheap) servers (eg. ppro or p2, 512 megs of RAM, a couple of decent SCSI disks, couple of NICs, etc.) then had some massive load balancing hardware in front of them, you could probably do some very cool things in terms of total cost versus performance. Finding a load balancing box to handle that sort of traffic properly could take some doing and you need a tech from the vendor there to help with issues with the load balancer, but you may be able to get that without too much trouble since it could be great advertising for the company involved. This approach truly exploits commodity PC hardware, the lack of licensing costs for free software, etc. But again, approaching this from the viewpoint of making a media circus before you have any results dooms you to failure. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-chat" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.3.95.980426140131.12990B-100000>