When it comes to High Performance Computing (HPC), it is still a world of haves vs. have-nots for all practical purposes. And it does not matter whether you work for a Fortune 100 company or an upstart startup You either have an HPC cluster when you need it or you don’t. It is as simple as that. If you are among the haves, then you get to test your ideas and run your simulations on the zealously guarded, well utilized, and efficiently operated HPC infrastructure that was “state of the art” when it was procured. The thought of asking for a refresh to the latest CPU and some GPUs does cross your mind, but you don’t act on it, for you know very well how that conversation ends. If you are among the have-nots, you wait in queue for your chance to test your latest idea or to run your next simulation. But, it does not have to be that way.
There is a solution to this supply and demand issue that will make both HPC users and the HPC infrastructure owners happy. Move your HPC workloads to the cloud. If you are a user of HPC infrastructure, you can get instant access to the HPC infrastructure of your choice, and the latest technologies by leveraging HPC clusters in the cloud. If you are an HPC Infrastructure owner, leveraging the virtually unlimited HPC resources in the cloud can finally help you meet the backlog of demands. You no longer have to be the gatekeeper who always says, “get in the queue.”
There can be no better example of this than the recent announcement from Descartes Labs that it had nabbed the 136th spot on the top 500 list of the world’s fastest publicly known supercomputers – with $5,000 and an Amazon Web Services (AWS) account. You can read this amazing story here.
One of the biggest advantage of moving HPC workloads to the cloud is the freedom to get started when you want it, and with the resources you want. The days of submitting your request for time on a cluster, months in advance, and then adjusting your development / research schedules based on the allocation you receive can be a memory of the past. If you work for an organization that has moved HPC workloads to the cloud or has at least started the process by bursting to the cloud when demand spikes, you have experienced the agility and flexibility benefits afforded by the cloud. You either have an individual account to access and request resources in the cloud or you request it via your HPC admin. In both cases, you start building “your” cluster when you are ready. When the cluster is built, you submit your jobs. And when the jobs are done, you shut down your cluster and stop paying for it. When you request your cluster, unlike your old on-premises environment, you can specify what type of CPUs (or GPUs, or FPGAs) you would like to run a particular application on. Ever wonder how much faster your application would run if you had the latest CPU or GPU? Well, now you can try it out without going through a cumbersome procurement process. Then, there is the scale. It does not matter if you request 100 cores for 8 hours or 800 cores for 1 hour. You essentially pay the same. So, if your application supports it, why not scale up your resources and get to results faster? Western Digital did exactly that when they built a cloud-scale HPC cluster on AWS and used it to simulate crucial elements of upcoming head designs for their next-generation hard disk drives (HDD). Their simulation encompassed a little over 2.5 million tasks, and ran to completion in just 8 hours on a million-vCPU Amazon EC2 cluster.
HPC in the cloud is the democratization of high performance computing – giving researchers the freedom to ask the questions and explore the ideas they have always wanted to and infrastructure owners the resources to meet their users’ needs. HPC in the cloud is the great equalizer – lowering the barriers to entry for startups and small universities and schools to harness the power of advanced computing to advance science and engineering. At AWS, we have combined the latest compute, networking, storage, security, cloud orchestration, and visualization technologies with a vibrant partner and ISV community to offer you a highly customizable computing platform to run a broad range of compute intensive workloads. Take us for a test ride. Build your first HPC cluster on AWS.