Speed and Scale

icons_speed_01sm icons_pro_02sm icons_col_03sm icons_cost_04sm icons_servi_05sm icons_just_06sm

Application Performance, System Efficiency and Scale

Speed and Scale are core to enabling the capacity of a system to achieve the end results an organization seeks to accomplish. Scale helps solve bigger problems or process more of them, and speed helps get results faster. The following are examples of capabilities, which contribute to speed and scale at an application or system level, as well as elements, which improve overall system efficiency:

  • Proven At Scale
    Moab Manages Some of the Largest Compute Environments in the World
    Moab has been used on some of the largest computing environments in the world. They do/will include either being the first or one of the first systems of the following scale: 1, 2, 10 and 100 petafop systems.
  • Multi-Cluster/Grid
    Scale Resources and Balance Workload Management Across Multiple Systems at Multiple Locations
    Unify capacity of multiple clusters by consolidating workload management. Grids enable users, groups and projects to more easily share resource and data.
  • High Throughput Workload Management
    Accelerate Launch Time for Large Volumes of Small Jobs with Nitro – High Throughput Manager
    Nitro helps users submit thousands to millions of small tasks. Rather than submitting many small individual tasks, Nitro packages these many tasks into a group request and then launches the tasks up to hundreds of times faster than a traditional scheduler. It can service SOA workload requests.
  • Optimize Heterogeneous Clusters 
    Meet the Needs of Diverse Users and Applications with Advanced Resource Management
    As clusters are scaled up to meet the needs of multiple groups, inevitably the application requirements of those groups will require different resource configurations to optimize their application performance. Moab’s Advanced Resource Management capabilities and Node Allocation Policies, Node Set resource groupings, NUMA, and other policies which help maintain efficient utilization in heterogeneous resource environments.
  • Elastic Computing
    Dynamically Scale your Cluster with Additional Capacity
    Elastic Computing allows Moab to add or remove resources from either a local resource pool, public/private remote resource pool or from within a shared cluster. This capability includes Moab monitoring wait times for workload and then triggering OpenStack, VMWare, HP-CMU, Bright Cluster Manager, other provisioning tools or Cloud APIs (Amazon, etc.) to provision additional resources to be under the workload manager’s control.
  • NUMA
    Accelerate Application Performance with Memory and GPU-aware Job Placement
    Proper NUMA-aware placement of a job can improve run-time by as much as 250% due to better memory access and can improve GPU data transfer by as much as 300%.
  • GPUs/Accelerators
    Scale Application and System Performance
    Automatically detect and scale/accelerate applications with support for accelerators such as NVIDIA GPUs and Intel Xeon Phi’s (MIC). Automatically apply applications to accelerators based on templates.
  • Power Efficiency/Constraints
    Improve System Power Efficiency with Power Management / Green Computing
    Enable your cluster to perform within power limitation constraints using Moab’s power management. Moab is able to reduce the power state of idle nodes in order to reclaim unnecessary energy usage. It can also manage power utilized on a per Application basis using clock speed and P-state management.