An Enterprise Server Cluster employs a logical star topology - each cluster instance is connected to a central cluster instance called the Global Lock Manager (GLM).

An Enterprise Server Cluster is composed of a minimum of two Enterprise Server Cluster clients and a single Enterprise Server Cluster lock manager - the GLM. In practice, the number of Enterprise Server Cluster clients is determined by the volume of JES work required - jobs can be submitted to any members of a cluster.

If desired, an Enterprise Server Cluster can present a single interface for job submission. Jobs can be submitted to a specialized instance of a cluster member - a scheduler region, which can itself execute JES work as well as orchestrate it. It is also possible for the GLM to act as a scheduler region - but we strongly advise (although do not enforce) that no work other than global locks processing is executed by the GLM. For example, it is inadvisable to create initiators on the GLM machine.

Current limitations:

  • If you submit jobs to a region that has no initiator, the first Enterprise Server Cluster server that requests work from the Job Manager for the target class will execute the work. This implies latency at least equal to the difference in time between when the relevant casmgr process requested the job from the Job manager, and the time at which the JCL was submitted. This latency will be addressed in a future release.
  • It is only possible to define a single GLM.