Hello
As a newcomer to the world of openMPI who has perused the faq and
searched
the archives, I have a few questions about how to schedule processes
across
a heterogeneous cluster where some processes must be bound to particular
nodes and where some processes will not be compute bound, meaning more
processes than CPU cores is likely desired. Allow me to first
describe our
situation and then ask some specific questions.
We have a cluster of four nodes/hosts. Two of these have a single
Pentium3
0.8GHz processor each. The other two have dual core processors. Our data
resides on two external drives that are mounted on the two fast machines
but are visible to all nodes over nfs. Our setup is not a real
cluster in
the sense that we are using ssh over nfs to link our nodes.
For our application, some processes (manager and the datastreams) won't
take much CPU power while others will. Our application first assigns the
non-compute bound processes and then the high CPU power ones. Clearly we
would like to be able to assign the former processes to our two slower
machines. How can we get MPI to achieve this? What is the best way to
use
slot and max-slot commands in the hostfile to optimize performance? How
does max-slot behave when some node is listed more than once in the
hostfile?
Is there any way to control where processes are deployed apart from the
basic bynode or byslot round robin schedule?
Many thanks for your time.
Regards
Roo
