Many different industry organizations use HPC to solve their most complex academic, scientific and business problems. However, the cost of using HPC effectively is very high. In order to process a large amount of data, HPC clusters need powerful computing power, high-performance memory and storage, and low-latency networks. Some organizations run HPC workloads by building infrastructure locally, which requires high upfront investment, including lengthy procurement cycle and continuous management overhead (such as monitoring software and hardware updates). When the infrastructure is outdated and must be upgraded, it will inevitably face the challenge of limited flexibility. Many industry customers choose to run their HPC workloads in the cloud, taking full advantage of the security, scalability and flexibility provided by the cloud. Many engineers, researchers and scientists run their largest and most complex HPC workloads through Amazon cloud technology, and choose to use Amazon EC2 instances with enhanced networks (such as C5n, R5n, M5n and C6gn) to expand high-performance tightly coupled HPC workloads, which require high-level communication between instances and handle thousands of interdependent tasks. Although the performance of these instances can meet most HPC use cases, as the problem to be solved becomes more and more difficult and the workload continues to expand, the scale may increase to tens of thousands of servers in Amazon cloud technology, and customers hope to maximize the cost performance when running HPC workloads.
The new Hpc6a instance is designed to provide the best cost performance for running Hpc workloads on a large scale in the cloud. The HPC6a instance improves the performance-to-price ratio of HPC workload by up to 65%, and can perform complex calculations in a series of cluster scales (up to tens of thousands of cores). By default, the Hpc6a instance is equipped with a Network Interface Elastic Fabric Adapter (EFA). EFA network has low latency, low jitter and network bandwidth as high as 100 Gbps, which can help customers improve their operation efficiency and deliver the calculation results quickly for the workload of massive communication between instances. The Hpc6a instance is equipped with the third generation AMD EPYC processor, which runs at 3.6 GHz and provides 384 GB of memory. With the example of Hpc6a, customers can solve their biggest and most difficult academic, scientific and commercial problems through Hpc more economically and efficiently, and get many advantages of cost-effective Amazon cloud technology services.
David brown, Vice President of Amazon EC2, Amazon Cloud Technology, said, "By continuously innovating almost all types of workloads and launching a new purpose-built Amazon EC2 instance, we have provided our customers with super high cost performance for some of their most critical business applications. Although high performance computing has helped solve some of the most difficult problems in science, engineering and business, it is still very expensive for many organizations to run HPC workloads effectively. The Hpc6a instance is specially built for Hpc workloads, which can help customers improve the performance-to-price ratio by up to 65% in HPC clusters of almost any size, eliminate customers' concerns about cost, and focus on solving core business problems. "
Dan McNamara, AMD's senior vice president and general manager of server business, said, "We are very pleased to continue to cooperate with Amazon Cloud Technology to provide their customers with a new and powerful example of high-performance computing workload. AMD EPYC processors are helping customers of all sizes solve some of the most difficult and complicated problems. From universities to enterprises to large research institutions, the HPC6a instance supported by the third-generation AMD EPYC processor brings powerful HPC performance and cloud scalability to more global customers. "
Customers can use the Hpc6a instance through Amazon ParallelCluster (an open source cluster management tool), preset the Hpc6a instance with other instances, and flexibly run different optimized instances for different workloads in the same Hpc cluster. The example of Hpc6a is based on Amazon Nitro system. Amazon Nitro offloads many traditional virtualization functions to special hardware and software, which has higher performance, higher availability and higher security, while reducing virtualization overhead. An instance of Hpc6a can be purchased as an on-demand instance or a reserved instance, or through a savings plan. Examples of Hpc6a have been launched in the eastern United States (Ohio) and Amazon GovCloud (western United States), and other regions will be launched soon.
Maxar is a geospatial information company, which cooperates with many innovative enterprises and more than 50 governments to monitor global changes, provide broadband communication, and provide advanced space operations based on space architecture and geo-intelligence capabilities. "The launch of Amazon EC2 Hpc6 instance is another major release of Amazon cloud technology, which enables Maxar to continue to meet and exceed customers' requirements for large-scale computing workflow, whether it is to accelerate the research and operation of numerical weather forecasting workload or to create the world's best, advanced and accurate digital twin model using Maxar Precision3D product suite. "Dan Nord, senior vice president and chief product officer of Maxar Technologies, said," The combination of AMD EPYC (Milan) processor of Hpc6a and EFA network function helps us to improve the performance by 60% compared with other options, and it is more cost-effective. This allows Maxar to make strategic choices in a set of HPC cluster configuration of Amazon cloud technology built by itself, to meet the needs of customers to the greatest extent, and to maximize flexibility and flexibility. "
DTN's global weather station network provides ultra-local, accurate and real-time weather information, and provides operational insights for enterprises. Lars Ewe, chief technology officer of DTN, said: "The cooperation with Amazon cloud technology enables us to better serve our customers, provide them with high-resolution weather forecasting systems and provide data for analysis engines. We are very happy to see the high performance-to-price ratio of the Hpc6a instance, and hope that it will become the first choice Amazon EC2 instance for our future Hpc workload. "
TotalCAE has more than 20 years' experience in computer aided engineering (CAE) high performance computing. TotalCAE helps customers to eliminate IT problems by managing their HPC engineering environment and engineering applications, so that customers can focus on engineering instead of IT. "The TotalCAE platform allows CAE departments to easily gain the agility and flexibility of Amazon cloud technology, and it can be used in hundreds of engineering applications with just a few clicks, such as Ansys Fluent, Siemens Simcenter STAR-CCM+ and Dassault Systèmes Abaqus." Rod Mach, president of TotalCAE, said, "As an HPC capability partner of Amazon Cloud Technology, we help customers run their CAE workloads in the cloud. Through the example of HPC6a, we have improved the performance of computational fluid dynamics workload by 30% at a lower cost, enabling TotalCAE to provide customers with industry-leading cost performance and strong cloud scalability. "