[cobalt-users] OT ish - real world cobalt server "serving" capacity comparisons.

["BT" <error404@xxxxxxxxxxxxxx>]

This is something I have looked for now and again, and never really found,
and now it has re-surfaced with a different customer request...

the basic question could be answere if there was a table for each raq with
one axis (basic html / php / php+mysql ) and the other access (server load /
10% / 20% ~~~ 90% / 100%) and the central table cells themselves contain "n"
where n= maximum concurrent users...

I know there are various "statements" that say raqx can handle up to a
million html pages a day, but on the other hand I have seen raqs stagger
under 15000 php / mysql pages per day...

http://www.acme.com/software/thttpd/benchmarks.html is nice but it doens't
answer my questions...  I guess specweb99 might give me the answers
(http://www.specbench.org/osg/web99/results/web99.html) but no numbers for
cobalt, but it all seems to list (proper) xeon configurations, not weedy
cobalt cpus....

I am hoping that some of you may at some time already crossed this bridge
and have these answers, because at the moment nobody (here) has any idea
whether the application in question, which is in main three separate php /
mysql systems under one domain, will cope with the anticipated load on a
single raq4, or whether we should stick each database "system" on a separate
raq4 now and be done with it... the latter is of course the safe option but
nobody is going to be very impressed if we do this and cpu load never goes
above 5%... the trouble with raq's is there is no way to stick in a MEC + 4
Gb RAM and an extra processor when things get treacly...

basically I am trying to determine if cobalt is even the right platform for
this sort of thing, or should we go out and buy a 2U or 4U "proper" server
for supermicro or something, cobalts have the advantage because they can be
added to the cabinet and running literally within an hour, and they are
"familiar technology" to the client...

any ideas / suggestions?