Re: caching web traffic

["B.C.J.O" <fade@deepsky.com>]

On Tue, 25 Jan 2000, Bulent Murtezaoglu wrote:

> >>>>> "BCJO" == B C J O <fade@deepsky.com> writes:
> 
>     BCJO> ... The NetApp worked admirably. I see no reason why
>     BCJO> squid on an Alpha wouldn't do just as well, though, at a
>     BCJO> fraction of the price. ...
> 
> It depends on what kind of load you put it under AFAIK.  There is some
> banchmark data at:
> 
> http://www.ircache.net/Polygraph/Results/

Thanks for the pointer; I looked over this data /w interest. Their numbers
seem to underline the kind of performance I was seeing on the
NetAppliance. I've found clusters of squid boxes to be a very flexible
solution for this type of reverse caching scenario -- as the data in your
link suggests they respond well to a fast disk subsystem /w many
spindles. Some of the same criteria you would apply to an nntp system
bears out in this case as well. =) It was also interesting to see the
non-linear cpu-utilisation graph. Obviously cpu throughput falls off as
a cache fills, but this spike will be cyclical through the life of an
operating array.

Brian
fade@etc.etc.etc.

How often I found where I should be going only by setting out for
somewhere else..
                                       -- R. Buckminster Fuller