[Beowulf] KVM to a compute node - ssh
Robert G. Brown
rgb at phy.duke.edu
Thu Jun 3 10:59:11 PDT 2004
On Thu, 3 Jun 2004, Michael Will wrote:
> This is a component we put into our products, rather than a standalone
> If you look at a 1U opteron system for example, you can add it for $21 in the
> "additional network interface" section.
No kidding! Gee, I missed this when ordering my stack of 1000Es. I
suppose that it is an Altus-specific product, then. And curiously
enough, it is two Penguin 1000Es of mine that are stuck right now and it
was them that I tried to reboot with Don Becker's WOL tool (and failed).
Any hints on rebooting a hung opteron via WOL and the network? Can a
1000E be configured so that this will work, or will WOL only work to
turn one on if they are off?
You see, a 32 port terminal server and the $21 interface and the cable
per node cost more than $100 per node to be able to remotely power cycle
it. That is more than the cost of an extra 1000E itself. We'd much
rather buy an extra 1000E than buy the TS and interface (especially
since we'd have to buy three or four TS's to cover all our nodes, which
starts to look like real money even if a mere $3K doesn't), but we'd
really like to have this cake and eat it too -- still be able to
remotely bounce the nodes -- and not have yet another box that can break
and that eats resources. In the meantime, $3200 also pays for a lot of
walks downstairs, regardless of the inconvenience.
If the server room were the other side of campus (and hence cost thirty
plus minutes of time to walk over and reboot instead of five), things
might be different, but in all economic decisions there are crossover
points where it isn't worth it to buy something that is really pretty
expensive just for the convenience.
Robert G. Brown http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567 Fax: 919-660-2525 email:rgb at phy.duke.edu
More information about the Beowulf