On Thu, 12 Apr 2001, Joseph Mack wrote:
> I had assumed (incorrectly) that the test setup would simulate a
> relatively few clients connecting through an L4 switch to web-caches
> that were fetching pages from a large number of websites (the whole
> internet).
This is a proxy load balancing setup (except you still want to have
lots of clients). We are inclined to use server load balancing setup
for the first switch-off because it is simpler and, according to some,
has significantly larger market.
> So there is no decision on the number of unique web-sites that the
> servers will be generating, or on the level of duplication? This
> will be decided by the participants? (I can handle either, I just
> need to plan for it).
There is no final decision, and we will want to hear participant
feedback/opinions before making one. The simplest approach would be to
use one VIP, one "web site", and appropriate number of simulated
servers behind that VIP. My personal preference for the first round of
tests is to make the workload as simple as possible while keeping it
realistic and generally useful.
> What OS/version will the servers be running and approx what
> hardware will they be running on
I expect the clients and servers to be similar to what was used for
the third cache-off, perhaps with beefier CPUs.
http://www.measurement-factory.com/results/public/cacheoff/N03/report.by-meas.html#Sect:5.3
I would expect each client-server pair to be able to generate about
700-800 HTTP transactions per second (about 60-70 Mbits/sec in
response traffic). We will post details and some results soon. Please
stand by.
Alex.
This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:18 MDT