A Place to Explore the Forefront of Compute: Introducing Memory-Driven Computing Sandbox

[Sean_Sargent]

By Hande Sahin-Bahceci, AI, Analytics, Security, Emerging Technologies Marketing, HPE Pointnext

Featuring Sean Sargent, HPE Chief Architect

If you were at HPE Discover 2018 Las Vegas in June, you surely experienced the excitement around Memory-Driven Computing, one of the major themes of the event. Memory-Driven Computing is nothing less than a complete re-visioning of the classical, processor-centric architecture of computing, enabling the processing of vast amounts of data in an in-memory environment and driving massive gains in performance. HPE announced the launch of the Memory-Driven Computing Sandbox and an incubation practice with special skills for high performance data applications, leveraging the expertise of HPE Pointnext and Hewlett Packard Labs. After Discover, I caught up with Memory-Driven Computing expert Sean Sargent, chief architect with HPE, to learn more about the Memory-Driven Computing Sandbox and what it can do for businesses.

Hande Sahin-Bahceci: Can you join up the dots between the announcement at HPE Discover Las Vegas 2018 and HPE’s broader strategy for Memory-Driven Computing?

Sean Sargent: The Memory-Driven Computing Sandbox is essentially a way for companies to explore this new technology. HPE Pointnext is engaging with a number of customers in early trials around Memory-Driven Computing, and we decided that a good way to accelerate these engagements would be to work with Hewlett Packard Labs to create this new incubation practice and development environment. It’s a way for people to test their own workloads and see how they can get to the kind of 10x and 100x speed-ups that we’ve helped other customers achieve.

Not only do they get access to the environment, they get access to new experimental features such as Software-Defined Scalable Memory, which enables our memory fabric to address larger pools of shared memory than ever before. It lets you compose memory and present it to multiple OSs. Plus, Sandbox users get early access to other innovations coming out of Hewlett Packard Labs.  I want to emphasize that we’re not just emulating these components – what we’ll have available in the Sandbox environment is the actual physical infrastructure.

The goal is to learn from our customers, as well. There is the opportunity to co-develop and collaborate on Memory-Driven Computing technology, but we are also trying to understand what advisory and operational services needs our customers will have in adopting Memory-Driven Computing. That's why we describe it as an incubation practice. There’s an element of collaboration and experimentation here that makes this really exciting.

Sahin-Bahceci: What does the Sandbox environment look like?

Sargent: We’ll be hosting on some extremely large systems. Customers don’t have to invest in having high-end systems on premises – we provide the raw horsepower. The Memory-Driven Computing Sandbox will be a 64 socket system on HPE Superdome Flex servers, with 48 terabytes of shared memory available to it. It will be located in Roseville, California, and select customers can access it remotely. It’s designed to get you started quickly with Memory-Driven program development, because the systems are already configured and up and running, with all the APIs you need, in a familiar Linux environment.

Sahin-Bahceci: What are some ways that companies might want to leverage the Memory-Driven Computing Sandbox?

Sargent: Workload performance testing is a big one, of course, but bear in mind, this isn’t for your average-day workload. Memory-Driven systems really come into their own when you want to optimize very complex, high-end workloads with massive data sets. We saw that working brilliantly at DZNE, for example. [DZNE is a German research institute that’s using HPE Memory-Driven Computing in the battle against neurodegenerative diseases; see my post Memory-Driven Computing: The Perfect Answer to Compute’s Perfect Storm.] We’ve also seen a lot of interest from software providers that want to see how their products stand up under massive workloads.

Then there’s big data/real-time types of applications. A lot of data-heavy online organizations, for example, are interested in accelerating their analytics to reduce the latency between a customer entering an online request and the results being displayed on the screen. If you can reduce that period, you lose fewer customer contacts, you've obviously got greater throughput because you can service more queries, so you increase your revenue.

There’s tons of potential use cases, and I’ve written about some of them. [See 3 Things You Can Do Better and Faster with Memory-Driven Computing]. One that I came across recently is security threat intelligence. If you think about the number of logs generated in an enterprise IT operation – imagine if you could store all of those in memory and query them at will. You'd have a much better understanding of your security posture, not just from an audit perspective after something goes wrong, but in real-time. You’d be in a much stronger position to take action if a new threat is detected.

Sahin-Bahceci: Any customer successes that come to mind?

Sargent: At Discover, we talked a lot about Travelport, a U.K.-based travel commerce platform that processes more than USD 80 billion in travel spend per year. They connect travel providers with online and offline buyers, and it takes enormous amounts of data, as you can imagine, so it’s a great fit for Memory-Driven Computing. They’re a longtime HPE customer. In collaboration with HPE Pointnext and Hewlett Packard Labs, we’ve been working with them to increase processing capability and make more memory available so their application can increase performance and scale better, essentially – helping them optimize and tune their code to run in that environment. HPE Pointnext is also helping them get a clear picture of the economic side of it, the cost-benefit analysis.

I think it’s fair to say they’ve been very pleased with the results. It’s going to be exciting to meet more companies that want to use the Sandbox to explore the same kind of innovative moves that we’ve seen at Travelport.

Learn more about Memory-Driven Computing, and get started with HPE Pointnext today.