RE: Cacheoff results published.

[cut]

> Our box was a Thunderbird 800, with 3 7200 RPM
> disks, and 512MB of RAM. We couldn't handle more than 30 reqs/sec on
> hardware that size. We've got our work cut out for us, eh?

I am quite puzzled though, those results strike me at being quite low.

I am running my production server on a Linux double PIII-500 w/ 5x9Gb SCSI
reiserfs-formatted diskd-managed disks, 384 Mb of RAM, and at 50% (of only
one)
CPU load I can handle easily 85 reqs/sec, including NTLM authentication and
authorization-checking against a 1700-users ACL (and the code
to handle that sucks, I'm about to rewrite it as per Henrik's suggestion
unless somebody beats me to it).

This is a real-world situation, which might be different from others'
since we have a pretty good upstream bandwidth and thus our FD-sets are
usually pretty small (never seen more than 500 busy FDs).
Also, a noticeable percentage of those hits are 407/auth-denied's needed
to authenticate the clients (that are lighter than a 200/OK to manage)

This said, I would expect to handle about 150 reqs/sec before I hit
the limit and have to start thinking of ways to get to use the second CPU.

-- 
	/kinkie