Moab Workload Manager

1.0 Philosophy

A scheduler's purpose is to optimally use resources in a convenient and manageable way. System users want to specify resources, obtain quick turnaround on their jobs, and have reliable resource allocation. On the other hand, administrators want to understand both the workload and the resources available. This includes current state, problems, and statistics—information about what is happening that is transparent to the end-user. Administrators need an extensive set of options to enable management enforced policies and tune the system to obtain desired statistics.

There are other systems that provide batch management; however, Moab is unique in many respects. Moab matches jobs to nodes, dynamically reprovisions nodes to satisfy workload, and dynamically modifies workload to better take advantage of available nodes. Moab allows sites to fully visualize cluster and user behavior. It can integrate and orchestrate resource monitors, databases, identity managers, license managers, networks, and storage systems, thus providing a cohesive view of the cluster—a cluster that fully acts and responds according to site mission objectives.

Moab can dynamically adjust security to meet specific job needs. Moab can create real and virtual clusters on demand and from scratch that are custom-tailored to a specific request. Moab can integrate visualization services, web farms and application servers; it can create powerful grids of disparate clusters. Moab maintains complete accounting and auditing records, exporting this data to information services on command, and even providing professional billing statements to cover all used resources and services.

Moab provides user- and application-centric web portals and powerful graphical tools for monitoring and controlling every conceivable aspect of a cluster's objectives, performance, workload, and usage. Moab is unique in its ability to deliver a powerful user-centric cluster with little effort. Its design is focused on ROI, better use of resources, increased user effectiveness, and reduced staffing requirements.