Reduce Costs while Accelerating Data-intensive HPC Workloads

Print Friendly, PDF & Email

While high-performance computing (HPC) helps solve the world’s toughest problems, it doesn’t come without complexity. 

In the process of deploying HPC infrastructure, organizations need to consider many variables to ensure workload deadlines are met, innovation is empowered, and costs are managed.

On-premises systems are the traditional method for running HPC workloads. Yet there are downsides to on-premises HPC deployments you shouldn’t ignore. 

On-premises HPC has inherent capacity constraints limiting your agility and potentially slowing innovation. High capital expenditures for on-premises systems find organizations locked in the same infrastructure for years, even when performance becomes suboptimal. There is also a question of capacity usage, as sometimes organizations over or under plan capacity and workload demands can change over time. 

Infrastructure constraints create long queue times for running workloads. Your ability to bring products and services quickly to market is hindered.  

There is a way of achieving insights faster without breaking the bank.

Cloud infrastructure purpose-built for HPC

High-performance processors, memory, storage, and inter-node communication are essential for running data-intensive HPC workloads.

For instance, design engineers use finite element analysis (FEA) to simulate a range of physical properties, including structural integrity and how an object responds to vibration, heat, and flow. FEA is widespread in many industries, such as automotive, aerospace, manufacturing, and even healthcare.

To effectively run FEA, engineers and scientists require a large amount of memory and local storage to accommodate a high number of elements that need to be tested. The choice comes down to either increasing the memory and storage of one machine or deploying multiple compute instances that can run in parallel. By spreading the workload over multiple instances, you can increase the number of elements you’re testing and subsequently achieve higher simulation accuracy.

Still, running simulations over multiple instances effectively needs fast inter-node communications. There is no point in running nodes in parallel very fast if engineers need to wait for the inter-node communication to finish before proceeding with the next step. The faster the instances run in parallel and communicate together, the quicker the job turnaround time. 

This brings us to the question: How can you access scalable compute resources with high memory performance, fast interconnect speed, and local storage while keeping costs in check?

The AWS and Intel advantage

Amazon EC2 Hpc6id instances, powered by 3rd Gen Intel® Xeon® Scalable processors, bring the power of HPC within reach for organizations of every size and industry. 

In the example of running FEA, these instances empower engineers and scientists to run more complex and detailed simulations involving larger datasets. Amazon EC2 Hpc6id instances deliver up to 2.2X better price-performance over comparable x86-based instances for data-intensive HPC workloads, such as Finite Element Analysis (FEA). The result? Improved product quality, shorter time-to-market, and reduced product development costs.

AWS provides a broad range of scalable, flexible infrastructure solutions organizations can choose to adapt hardware resources to individual HPC workloads. You can choose the most appropriate mix of resources for your specific applications while ensuring costs are managed. You don’t have to worry about HPC maintenance, because AWS does the undifferentiated heavy lifting for you. 

Additionally, with flexible pricing and pay-as-you-go infrastructure, capacity planning worries can become a thing of the past. With HPC on AWS, organizations can flexibly tune and scale their infrastructure as workloads evolve instead of the other way around. 

Final thoughts

AWS puts the advanced capabilities of HPC in reach for more people and organizations while simplifying processes like management, deployment, and scaling. 

Accessible, flexible, and cost-effective, AWS lets organizations unleash the creativity of their engineers, analysts, and researchers for faster and better results.

Learn more and get started at Amazon EC2 Hpc6id Instances – Compute – Amazon Web Services