Compute nodes from one cluster must be able to send/recv
TCP/IP messages (i.e., what I understand you to mean by "see")
directly to/from compute nodes to another cluster. In short you
need to be able to open a socket connection directly between
compute nodes on two different clusters.
Nick
On Jun 15, 2007, at 5:00 AM, Mehdi Sheikhalishahi wrote:
Does Computing Nodes of a Cluster under their headnode need to see
Computing Nodes of another Cluster under their headnode or only
two headnodes need to see each other in mpich-g2 If in our rsl we
specify these two headnodes that in both of them Globus is
installed and theirGatekeeper configured? For example consider the
following topology:
LNode1-------
----------RNode1
LNode2-------
----------RNode2
LNode3-------
----------RNode3
LNode4-------
----------RNode4
LNode5-------LHeadnode
()--------------------------------------------------------------
RHeadnode()---------RNode5
LNode6-------
----------RNode6
LNode7-------
----------RNode7
LNode8-------
----------RNode8
--
Best Regards,
S.Mehdi Sheikhalishahi,
Web: http://www.cse.shirazu.ac.ir/~alishahi/
Bye.
