Adaptive Computing - Moab Cloud
Collectively Moab Cloud and the open-source TORQUE resource manager provide an intelligent workload-driven solution that delivers advanced policy management, scheduling and reporting tools for many of today’s most advanced systems.
Moab Cloud and TORQUE, provide the ability to schedule GPGPU jobs to run exclusively and finish in the shortest time possible for certain types, or classes of jobs or users to ensure job speed and success. The workload management system uses policies to further manage optimal placement across available nodes to maximize utilization. Policies are based on current real-time metrics that can affect the speed and success of GPGPU jobs such as utilization, total memory to ensure job needs are met.
- Favor GPGPU jobs over non-GPGPU jobs for GPGPU-equipped resources
- Intelligent packing of GPGPU jobs
- Distribute GPGPU resources, based on policies, among users, groups, and workloads
- Designate a preference for priority scheduling of various GPGPU types