From: David Jones <jones@mc...> - 2003-05-14 19:53:56
My understanding is that no one has had much trouble patching their host
kernel with skas3. But can anyone see any potential problems of patching
350 host kernels which are regulary used in a cluster environment for a
variety of scientific apps?
There might be a point where we try some scalability tests with uml on our
supercomputer. I have yet to talk our sys admins about it, but I can
already see some difficulties trying to use uml (root priviledge issues:
adding tap device, modifying routing tables, etc). It seems that asking
them so use an expiremental patch might be a bit much. Can they be
convinced that the ska patch makes minimal changes to the host kernel?
I'm only at the phase of testing between 2 computers and I'm moving to ~30
in the near futures. 350 machines will be some time off.