On Sunday 13 April 2008 15:31, Bob Lee wrote:
> You might consider running the operations in their own VMs. You can
> reuse them, let them die, and start new ones as necessary.
I can't just let servers die and walk away from a computation. When a
user submits a problem, I have to execute until completion or until a
limit (user-specified or system-imposed) is exceeded and the provide a