Various industries have adopted or are in the planning and evaluation phase for using the cloud for HPC applications. Within the realm of technical computing, certain workloads are suited to a cloud-based HPC environment. Workloads could be considered either loosely coupled or tightly coupled. In each of the industries discussed in this article, multiple jobs submitted with different input parameters would be loosely coupled and not require a low-latency, high-speed interconnect, while a job that requires the use of multiple systems working in concert would be tightly coupled and need InfiniBand (IB).
This article describes the challenges that users face and the solutions available to make running cloud based HPC applications a reality. You’ll learn about different cloud computing models, potential economic savings and factors to consider when comparing an on-site data center with a cloud-based provider.
Cloud computing is changing the way that IT organizations operate and innovate. While moving enterprise type applications to a cloud service provider is progressing, technical computing has been lagging in this transformation. This is changing, as the technology that once sat in a protected data center is now available in the cloud.
Demands by users that are running applications in the scientific, technical, financial or research areas can easily outstrip the capabilities of in-house clusters of servers. IT departments have to anticipate compute and storage needs for their most demanding users, which can lead to extra spending on both CAPEX and OPEX once the workload changes.
High performance computing in the cloud just got a lot easier. Omnibond, the South Carolina-based company that provides development and support services for OrangeFS, has released CloudyCluster just in time for SC14. The new solution works in conjunction with OrangeFS to ease the burden of creating and maintaining a cloud-based HPC or Big Data infrastructure.
“For this big workload, a 156,314-core CycleCloud behemoth spanning 8 AWS regions, totaling 1.21 petaFLOPS (RPeak, not RMax) of aggregate compute power, to simulate 205,000 materials, crunched 264 compute years in only 18 hours. Thanks to Cycle’s software and Amazon’s Spot Instances, a supercomputing environment worth $68M if you had bought it, ran 2.3 Million hours of material science, approximately 264 compute-years, of simulation in only 18 hours, cost only $33,000, or $0.16 per molecule.