Configuring computers to be compute nodes for MPP (Windows)

Follow this procedure to configure Microsoft Windows computers to act as a compute nodes in a parallel processing cluster.

Procedure

  1. Run the IBM® InfoSphere® Information Server installation program on each compute node to install the parallel processing components on the node.
    See Installing parallel engine components on compute nodes (Windows).
  2. Set system environment variables on the conductor node and compute node. Also verify that a temporary directory exists on the C: drive.
    See Making system environment changes (Windows).
  3. Configure a common user account for the compute nodes to use.
    See Configuring a common user account for parallel processing compute nodes (Windows).
  4. Configure RSH by setting permissions in the hosts.equiv file.
    IBM InfoSphere DataStage® uses remote shell (RSH) to communicate between the different physical computers in a parallel processing cluster. See Setting remote shell (RSH) permissions (Windows).
  5. If your users are Windows Domain users, configure the MKS Toolkit Rexecd and Rshd daemons to handle non-local user names that do not include a domain component.
    See Configuring the MKS Rexecd and Rshd daemons (Windows).
  6. To run parallel jobs on remote Windows hosts, define a remote shell script on the conductor node.
    See Configuring a remote shell script for the parallel engine (Windows).
  7. Set up a method of distributing the transform libraries to the compute nodes.
    See Distributing transform libraries to compute nodes.
  8. After you have set up the compute nodes in a Microsoft Windows parallel engine configuration, test the configuration.
    You can test the configuration by using either an MKS Korn shell or the IBM InfoSphere DataStage client. See Testing the configuration (Windows).