On 11/10/2008 09:39, Ethan Dicks wrote:
On Sat, Oct 11, 2008 at 10:00:49AM +0200, Jochen Kunz
wrote:
> On Fri, 10 Oct 2008 13:38:10 +0000
> Ethan Dicks <ethan.dicks at usap.gov> wrote:
>
Well, at least
some of the SGI Origin machines had a LCD like this in
every node, diplaying various things like CPU utilization etc..
Interesting. The largest/newest SGI machines I ever got to play
with were Onyx and Crimson boxes.
Could you access the LCD from user-land processes, or was there
always a system monitor running on them?
It's not in every node -- a node is usually taken to mean a processor
board, with two R10K or R12K processors and some RAM on it, 4 boards to
a module, and two modules per rack. Each module can be used as an 8-CPU
system, and has its own local management unit, which can control the
power, reset line, etc. The module midplanes are interconnected with a
Craylink to make a larger system, up to 256 CPUs. The main rack in such
a system has an LCD working off a little 486-based controller running a
ROM-based monitor. It acts as a system monitor and overall manager,
being connected to the console serial lines of the management units of
the modules in the rack. Under Unix you can send messages to the 486,
and the kernel does this to display system utilisation.
--
Pete Peter Turnbull
Network Manager
University of York