That's a very interesting idea. It's kind of a variation on the idea of creating the VM and loading the requested OS when a job runs. I'm not sure which one is faster (always a question that people will ask)?

This shows my limited experience with virtualization, but when you suspend a VM, you would dump it to disk - correct? Then to restart the VM, you just restart it from the disk image? Not a bad idea. One sort-of drawback is that you need a disk in each node :) I'm a fan of diskless nodes since it increases uptime (you would be surprised what impact hard drive failures have on the uptime of systems - I wrote an article about his using data from CMU - it's amazing and it's totally indpendend of the disk type), saves money, saves power (a little). So it would be difficult to save the image to disk unless it's a centralized storage point.

Thanks for the idea! It's very intriguing!

Jeff

 out of  found this valuable. Do you find this valuable?