A.12 Case Study: Holistic Cluster Scheduling and Consolidation

Overview

A large resource organization provides industry leading computing access but currently only schedules compute nodes. They would like to extend their cluster scheduling umbrella over their network, storage, and visualization services. Through this action, they seek the benefits of full cluster monitoring and intelligent workload steering, automated notification and cluster recovery, non-compute resource reservation and priority allocation.

In addition, the organization would also like to consolidate user and service level management to a single organization wide database and allow single system image presentation of all available resources including supporting a single, multi-resource manager wide submission point and language.

Solution

Using multi-resource manager technology, Moab is able to monitor and schedule non-compute resources guaranteeing resource availability, steering workload around failures and towards less heavily loaded network and visualization resources, and allocating resources according to a single integrated site wide set of mission objectives.

Using Moab's identity manager interface, access portal technology, peer to peer grid, and job translation features, the organization can present a single command line or web based submission access point for all resources regardless of the underlying cluster location, cluster architecture, resource manager, or policies.