Execution servers

Execution servers provide the ability to execute resources stored within the repository. When a resource is included in a job for execution, the job step definition includes the specification of the execution server used for processing the step. The execution server type depends on the resource.

Execution servers currently supported by IBM® SPSS® Collaboration and Deployment Services include:

• SAS. The SAS execution server is the SAS executable file sas.exe, included with Base SAS^® Software. Use this execution server to process SAS syntax files.
• Remote Process. A remote process execution server allows processes to be initiated and monitored on remote servers. When the process completes, it returns a success or failure message. Any machine acting as a remote process server must have the necessary infrastructure installed for communicating with the repository.
  Note: The IBM SPSS Collaboration and Deployment Services Remote Process Server has a default thread pool core size of 16, which allows a maximum of 16 concurrent jobs to be executed on a single remote process server. Any concurrent jobs in excess of 16 must wait in the queue until the available thread pool has free resources. To manually configure the IBM SPSS Collaboration and Deployment Services Remote Process Server thread pool core size, add the following JVM option (with a user defined value) to the remote process server's startup script: prms.thread.pool.coresize=<user defined value>

  For more information regarding the start-up script, see the "Starting and stopping the remote process server" section in the IBM SPSS Collaboration and Deployment Services Remote Process Server guide.

Execution servers that process other specific types of resources can be added to the system by installing the appropriate adapters. For information, consult the documentation for those resource types.

The IBM SPSS Modeler execution server is IBM SPSS Modeler Server, which permits distributed analysis for data mining and model building. This execution server requires the specification of user credentials under which the processing occurs. See Server definitions for more information.

Two types of execution servers are available for processing IBM SPSS Statistics syntax files. The first is a remote execution server that corresponds to the IBM SPSS Statistics server product. This execution server requires the specification of user credentials under which the syntax processing occurs. The other type of execution server is local to the repository and corresponds to the batch facility included with the IBM SPSS Statistics server product. However, unlike the IBM SPSS Statistics server product, the batch facility can be run from the command line and requires no user credentials for execution.

To allow load balancing, two or more execution servers can be grouped together in a server cluster. When a job step uses a cluster for execution, IBM SPSS Collaboration and Deployment Services determines which managed server in the cluster is best suited to handle processing requests at that time. For more information, consult the IBM SPSS Deployment Manager documentation.

During job creation, assign an execution server or server cluster to each step included in the job. When the job executes, the repository uses the specified execution servers to perform the corresponding analyses.