Are Outdated Misconceptions About Cloud Infrastructure Depriving You of Its Benefits?

High-performance computing (HPC) was once considered the domain of research organizations and large enterprises with massive budgets for compute and storage. Now, however, organizations of many types, across many industries, are discovering the benefits of using the cloud to run tightly coupled, MPI (Message Passing Interface-based) HPC applications to harness the increasingly large amounts of data that fuel modern enterprises. Use cases are as varied as the data that organizations now have access to, ranging from computational fluid dynamics, weather prediction, and molecular dynamics to vehicle crashtesting, seismic processing, and astrophysics. These types of workloads have traditionally been run in onpremises data centers because of the compute power needed to run multiple processes simultaneously while regularly exchanging information along each step of the process. These workloads can be defined as tightly coupled, MPIbased HPC applications that depend on fast and efficient communication between proces and the ability of the underlying infrastructure to quickly distribute data among all the nodes involved in solving a particular problem. Cloud platforms, with their virtualized server environments, have long been considered unsuitable for tightly coupled, MPI-based HPC workloads. That perception is changing.