Dan's Take

Supercomputers-as-a-Service

Cray and Markley Group team up to extend the vision of cloud computing.

• By Dan Kusnetzky
• 09/20/2017

Adam Burnham, business development executive, Markley Group, and Ted Slater, global head of healthcare and life sciences, Cray, recently chatted with me about their recently announced partnership to provide  supercomputing-as-a-service solutions based on Cray supercomputers and the hosting capabilities of Markley. They also discussed their thoughts about  industry-specific solutions.

Why Supercomputing?
There are some computing tasks that require systems to munch through extremely large stores of data or require extremely complex computations based upon a much smaller amount of data. While some of these tasks can be performed using standard, off-the-shelf industry standard systems, there are some tasks that are so complex or have to deal with such huge amounts of data that a cluster or grid of industry-standard systems just isn't up to the task. It would simply take too much time.

The industry has addressed these needs with special-purpose computers that process data differently. These machines, such as the Cray Urija-GX that Markley Group deploys, have enhanced capabilities to deal with "chunks" or "vectors" of data, or have been enhanced to do graph processing. Graph processing stores data in nodes, and traverses these nodes while conducting "semantic processing."

Cray points out that it has been a long-time supplier (the Cray 1 was announced in 1975) of this type of special-purpose system.

Burnham and Slater noted that supercomputing capabilities have become increasingly desirable across a range of industries, including life sciences, bio-pharma, aerospace, government, banking and more. These organizations often have requirements to wade through extremely complex or very large data sets while conducting their research and analysis. Supercomputing makes it possible for them to do this type of work in a timely fashion.

Cray and Markley Group decided that if they worked together, they had the ability to offer these organizations access to supercomputing as a cloud service offering. They said their goal is making it easier and more affordable for research scientists, data scientists, and IT executives to access dedicated, powerful compute and analytic capability to increase time to discovery and decision, even when they don't have the budget to purchase them and set up a supercomputing datacenter.

Cray and Markley Group are making these capabilities available both on-premises and in the cloud.

Dan's Take: Under the Radar
There are many suppliers of cloud services for those interested in placing industry-standard, x86-based, workloads in the cloud. We hear stories about enterprises using these services in new ways almost daily. We seldom hear stories about other types of computing in the cloud. Why?

One reason is that the market is somewhat limited compared to hosting industry-standard workloads in the cloud. There are only a few cloud service providers offering mainframe-as-a-service or single-vendor-computing-environments-as-a-service, such as IBM i or  IBM AS/400 as-a-service. I've seldom come across supercomputers-as-a-service.

Such a service would be of great interest to organizations focused on research, engineering, geophysical analysis for oil/gas or other extremely computationally intensive tasks that are either trying to reduce their expenses when first trying out this type computing, or to reduce their expenses when they only have occasional need for this type of computational power.

Maybe it's the requirement for extreme levels of technical and industry-specific knowledge required, or maybe the limited audience for those services that have kept suppliers out of the market until now.

Whatever the reason, if your organization has a need for supercomputing but isn't interested in building its own internal tools to address those needs, or you're just getting your feet wet in this area, this joint effort of Cray and Markley is worth learning about.