On Thu, 28 May 1998, Tim Brody wrote:
> How about a graph of price/performance? :-)
>
> Whatever the figures that whoever can roll out, it would need an independent
> body of both squid or netcache to give genuine fairness of testing.
very true. Hmmm, anyone want to send me a netcache box to play with,
purely for comparision purposes you understand ;)
> Also, I should imagine the squid that they testing with was probably 1.1.x
> which has a far higher response time than 1.2, in my humble set up!
On looking at the rest of the thread (which has Stewart gone off into
depths I leave to coders ;) ), it would seem that (paradoxically), the
best way to 'scam' better performance out of your low-request rate
squid1.2, is to have a client purely doing (incache) requests to boost the
io-threads up to where they start working enough (assume your acls don't
let this request client generate any 'real' traffic).
Hmmm, speed the cache up by giving it more to do... dunno about that ;)
--==--
Bruce.
Received on Thu May 28 1998 - 18:53:38 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:40:30 MST