ホーム / Clouds / Platform ISF Adaptive Cluster / HPC Cloud

HPC Cloud

Platform ISF Adaptive Cluster

Use Case – HPC Cloud

Infrastructure Test/Dev Cloud
HPC Cloud Capabilities
  • Provides a self-service infrastructure
  • Re-provisions resources automatically
  • Harvest resources within the datacenter and/or desktops automatically
  • Allows “Cloud-bursting” to leverage external resources
platisf_hpccloud_uc.gif
HPC Cloud Scenario

HPC users need computing resources for running their HPC jobs. Typically these users are given a cluster of resources that are purchased and configured with the single operating system. If multiple operating systems are required then another cluster is created. Each cluster of resources are independent, however the users can have their needs met with varying requirements.

Problems that an HPC Cloud addresses

As jobs are submitted they can run on the respective cluster. However, each user may have different scheduling needs that can change over time from their users. One problem that can occur is when clusters are shared by users from other groups based on need. This creates resource utilization conflicts among user groups. Another problem occurs when a cluster cannot handle the processing requirements because there are insufficient resources, or added cost is required to increase the size of the cluster.

Solution offered by an HPC Cloud

All the HPC clusters can be combined into a single (larger) cluster. Users can continue to submit their HPC job but they do so through Platform ISF. Platform ISF will determine the resource requirements, and will dynamically provision the server resources to meet those requirements. While jobs are running on the cluster, Platform ISF Adaptive Cluster will also manage the workloads on the set of resources in order to optimally share the pool of resources for maximum HPC application throughput. Additionally, with the addition of Platform ISF Adaptive Cluster workload management policies can also dictate if and how an application can run workloads on external services. This enables the ability to burst to harvest internal resources such as underutilized datacenter servers and desktops, as well as outside pubic computing resources on demand.

Benefits of an HPC Cloud
  • Higher utilization of all servers
  • Smarter application on-demand usability
  • Scalability of higher cluster loads supported without adding new servers

;