Date: Sat, 25 Apr 1998 18:44:58 -0600 (MDT) From: Marc Slemko <marcs@znep.com> To: Don Wilde <dwilde1@ibm.net> Cc: freebsd-advocacy@FreeBSD.ORG Subject: Re: *** Real Action Item: SPECweb Message-ID: <Pine.BSF.3.95.980425183455.28001h-100000@alive.znep.com> In-Reply-To: <3542368D.C34933A8@ibm.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 25 Apr 1998, Don Wilde wrote: > If I read the results, we'll be able to match Novell's faster web > software with our faster OS and okay websoftware by using much faster > disks and equal net performance. As you know, my original Challenge used > $3000 hardware, and eventually I'd like to force them to compete on that > which is really _our_ turf. However, I think we can match or beat their > numbers with trick hardware. Even if Apache is slower, if our hardware > is faster and we solve the bandwidth limit problem so we can USE the > hardware performance we are designing, we won't just get 'ok' numbers. Again, you are underestimating the effort put into some of these results and the difficulty of getting results. > > Lastly, if you think 2.0 is going to be better and is going to be > released within 6 months of June 28th, why not help us by tweaking _it_ > to be better and faster. Just cut an interim release which is our > tweaked version, and that satisfies the 'generally available' clause. It is hard to cut an interm release of something that isn't written. > > I have gotten great benefit from your devil's advocate comments, and I > hope you will continue to contribute to this effort. That said, I hope > you can switch gears and start figuring out what you _can_ do to help us > 'win' in this attempt. We're going to do it, will you help? As I have said time and time again, if you are trying to beat existing numbers (as your above comment indicates you are still trying to do; your target isn't trivial since AFAIK Novell's results are currently the best single processor results listed) you must use Zeus to have any hope, and even then you may or may not succeed. I don't know how many ways I can say that you can't approach this trying to go in and kick ass or you will not end up with anything productive. I think that getting results for various under-represented solutions (in this case, Apache and FreeBSD) is a positive experience even if they are middle of the range. You do have to remember that most companies don't submit SPECweb numbers unless they are right up there, so even something in the bottom half of the published results can be a good thing. The attitude that I am still seeing, however, about what can be accomplished is a bit naive. Even better would be some way to establish a lab or (preferrably) a good relationship with a vendor with existing labs in different areas that can be freely used by free software vendors for benchmarking and performance tuning. There are parts of software that simply can't be tuned right for high load without being able to actually benchmark it under load, and those resources can be hard to find. I really don't have the spare time to get overly involved, but I am trying to caution you against going into this with unrealistic expectations and having a big flop. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-advocacy" 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.3.95.980425183455.28001h-100000>