Hello everyone,
I hope my post won't sound too inaccurate or isn't redundant (I haven't found much in the Linux community on the subject. I run a little GSP (Game Server Provider) company here in Europe and I am having difficulties to optimize my IT system for gaming.

One of the things I would like to know is how to monitor (is there a tool) each "core" of a CPU (we have a couple of Dual Cores and a bunch of Quad Cores) ?

The context of this question is optimization: I currently host about 15 gameservers / server and would like to use "core" load balancing (since from what I read, hlds and srcds don't support hyperhtreading) in order to make each "core" balance the total number of gameservers running on the CPU.

There are never more than 8 to 10 gameservers launched on a server at the same time, son I was wondering if I could force the CPU to launch 2 or 3 gameservers per core.

So basically:
- does anyone know if this is possible ? and if so, how ?
- is there a tool that can monitor each CPU "core" individually ?

BTW, I'm running Debian on all servers and gameservers are all on a big nfs server, processes are launched on different servers from there.

I thank all of you for the time you take to read and answer my questions.
Best regards,
Arnaud
eSport-Eu

cgroup (control group) containers will probably be your best bet. This allows you to isolate specific CPUs/cores to separate "containers", and assign workloads to those containers.
With (very) recent kernels you can also assign relative weights to containers - say 35%/65% if you have a preferred workload. See ../Documentation/cgroup.txt

Thanks, I'll look into it How is the application "load" balanced now (I mean with le linux (debian) out of the box ?

Mmmm - scheduler 101 ... must be something out there on the 'net.
Recent kernels use the CFS (completely fair scheduler) - the management of containers I mentioned above was added by Ingo recently. Have a read of this
I have no idea whether Debian have the requisite (kernel) options selected.