Recently, AWS announced the general availability (GA) of Amazon EC2 Hpc6a Instances. With these instances, the company expands its portfolio of HPC compute options.
Earlier in November 2021, the company pre-announced the Hpc6a as a new HPC-optimized EC2 instance, which delivers 100 Gbps networking through Elastic Fabric Adapter (EFA) with 96 third-generation AMD EPYC processors (Milan) cores that run at frequencies up to 3.6 GHz and provide 384 GB of memory. With the GA of the instance, customers can efficiently run compute-intensive, high-performance computing (HPC) workloads, such as computational fluid dynamics, reservoir modeling, weather simulations, and finite element analysis.
Customers can provision Hpc6a instances alongside other instance types using AWS ParallelCluster (an open-source cluster management tool), giving them the flexibility to run different workload types optimized for different instances within the same HPC cluster. Furthermore, the Hpc6a instances themselves from the AWS Nitro System, which means no cores are held back for the hypervisor, thus making all cores available to code.
By default, Hpc6a instances are enabled with EFA, a network interface for Amazon EC2 instance, which provides the customer with low latency, and low jitter. Furthermore, it offers up to 100 Gbps of EFA networking bandwidth to increase operational efficiency and drive faster time-to-results for workloads that rely on inter-instance communications.
In an AWS News blog post on the new instances, Channy Yun, a principal developer advocate for AWS, explained the need for Hpc6a instances:
Customers today across various industries use compute-optimized EFA-enabled Amazon EC2 instances (for example, C5n, R5n, M5n, and M5zn) to maximize the performance of a variety of HPC workloads, but as these workloads scale to tens of thousands of cores, cost-efficiency becomes increasingly important. As a result, we have found that customers are not only looking to optimize performance for their HPC workloads but want to optimize costs as well.
In the same AWS News blog post, Yun stated that the benchmark test at Siemens Simcenter STAR-CCM+ of am automotive CFD simulation showed that the Hpc6a scales up to 400 nodes (approximately 40,000 cores). In addition, it also could maintain roughly 100% scaling efficiency with the help of EFA networking. And lastly, Hpc6a instances are 70% less expensive than c5n instances, implying that companies can deliver new designs faster and at a lower cost when using Hpc6a instances.
In an AWS Press Release, David Brown, vice-president of Amazon EC2 at AWS, said:
While high-performance computing has helped solve some of the most difficult problems in science, engineering, and business, effectively running HPC workloads can be cost-prohibitive for many organizations. Purpose-built for HPC workloads, Hpc6a instances now help customers realize up to 65% better price-performance for their HPC clusters at virtually any scale, so they can focus on solving the biggest problems that matter to them most without the cost barriers that exist today.
Currently, the Hpc6a instances are available in the US East (Ohio) and GovCloud (US-West) Regions in On-Demand and Dedicated Hosting or as part of a Savings Plan.
