Reply To: Communicate and work with other nodes

[Arun]

Host machine is usually a separate machine (computer that requires RAM/CPU etc.) that allows the user to connect to the cluster and launch and submit jobs on the cluster. It is typically set aside and not used to run jobs on it directly. Host machine is separate from the cores available on a cluster. Host machine is where you interact with the cluster and write your scripts to manage jobs submitted to the cluster. If you have 80 cores/nodes available and you submit the job using all 80 cores, then the mesh will be divided into 80 partitions, and each core/node will be assigned one partition. Keep in mind that one of these nodes is designated as compute node 0 (first node connected to the host via a socket interconnect), which connects to the host machine to receive the host commands and distributes them to other nodes. Other nodes will be designated as compute node 1 to compute node 79. All of these compute nodes from 0 to 79 (total of 80) are connected with each other to exchange data as needed during the simulation using inter-process communication.