Re: Bake-off

jlauro@umich.edu wrote:
>
> > THE bottleneck of Squid is well-known -- Unix file system. We are working on
> > SquidFS to eliminate that bottleneck. As for the discussions, there were
> > plenty on squid-dev list. I think it's time to actually implement something.
> >
>
> The bake-off did not have any test cases with high end
> hardware for squid. I would like to see the results of squid
> with decent hardware. Because of the Unix FS, you really need
> 10000 RPM drives, and would be better off and cost less with 3
> 10,000RPM drives (6 10000RPM drives even better) then the 6
> cheap drives it was tested with in the bake-off. Also, a
> $3500 switch and a single gigabit port along with squid
> running on the gigabit port. Just to see how squid "scales"
> when you put decent hardware behind it. It will bring the
> cost closer to $10,000, but wil give a better idea of how
> squid "scales" compared to the others. A low end configuration
> of course should still be tested.... but I think the low end
> would be better off with 2 7600RPM IDE drives.

We have some high-end hardware here. I'm planning to replicate the
bake-off tests on some twin Xeon2/450 boxes. I need the numbers myself
for a report anyway.

D
(Nobody hassle me about this. I got plenty to do)
Received on Thu Apr 08 1999 - 02:00:52 MDT