Re: www.microsoft.com sure needs a lot of silicon

Ron Atkinson (AtkinsonR_at_detroitedison.com) wrote:

: Your calculations are based on traffic being constant over a 24 hour
: period,and we all know that this is not so. During business hours
: traffic is much higher than your calculation, during the middle of the

One machine on a busy site I get to watch (not Microsoft) has the peak load looking to be at about 4X the minimum load. Using my Mk I eyeball (with the optional polycarbonate resolution enhancers (glasses) installed :) on a PerfView network packets per second chart of the beast implies that the peak is about 2X the mean.

: Ewald de Wit <ewald_at_bitterling.leidenuniv.nl> wrote in article
: <33540CF5.4003AB1E_at_bitterling.leidenuniv.nl>...

So assume then that the peak is 1200 hits/second. There are a number of single system SPECweb results at http://www.specbench.org/ handling that many hits/s. (SPECweb96 is pure GET, no CGI, average of about 15KB per hit, YMMV)

: > So each PentiumPro has to cope with an
: > average of 10 hits per second. Using one
: > PPro for serving a lousy 10 hits per second
: > is pretty pathetic, by all standards.

I've seen an NT IIS system do much better than that on SPECweb, which would make me wonder what the network connectivity was like to each system. A 10BaseT will max-out at about 60 SPECweb96.

rick jones
these opinions are mine, all mine; HP might not want them anyway... :) Received on Fri Apr 18 1997 - 00:00:00 CDT