AWS has unveiled instances that allow customers to run High-Performance Computing (HPC) workloads. The so-called Amazon EC2 Hpc6a instances feature the latest AMD EPYC processors.
Amazon EC2 Hpc6a instances are designed to solve complex academic, scientific and business problems. AWS claims that the instances can accelerate the processing of HPC workloads by a factor of 65 compared to other compute-optimized instances. Speed isn’t the only cost-saving benefit. According to Amazon, Hpc6a allows users to save significantly on computing power, memory, storage and low-latency networking requirements involved in deploying HPC clusters.
The Hpc6a instances are based on third-generation AMD EPYC processors. The processors deliver up to 394 GB of memory capacity with 100 Gbps of Elastic Fabric Adapter networking. This supports cost-effective scaling of HPC cluster deployments. According to the cloud giant, scalability extends to tens of thousands of cores. Such scale facilitates research projects in fluid simulation, genetic and molecular research, weather forecasting, seismic research, financial risk modelling and machine development.
Support from other AWS services
Hpc6a instances are supported by a number of specific AWS services. For example, AWS ParallelCluster. This cluster management solution allows the HPC instances to be deployed alongside other instances in the same cluster.
Amazon Nitro System supports the instances as well. Nitro System features building blocks for offloading traditional virtualization functionality to dedicated resources for better performance, availability and security.
The Amazon EC2 Hpc6a instances are available on-demand, as spare instances and through the AWS Savings Plans. The instances are available in the AWS US East (Ohio) and AWS GovCloud (US-West) regions. According to the cloud giant, support for other cloud regions will follow shortly.