HPE will be driving their Memory-Driven Computing model much more broadly to customers, partners and ISVs, by making a sandbox version available, particularly to expose it to developers, and also by bringing it to market selectively through their Pointnext flexible consumption model services.
The panel: Mark Potter; Kirk Bresniker, Jeff Kyle, Balaji Subramaniam Venkatesan, Matt Minetola and Dr Sharad Singhair
LAS VEGAS — A long time ago [2014] in a company far, far away [the old HP] what became Hewlett Packard Enterprise unveiled The Machine, a new computing model to decouple memory from processors and ramp up dramatic performance gains, in what today is known as Memory-Driven Computing [MDC]. This week, at the same venue in Vegas where The Machine was introduced at the 2014 HP Discover, an HPE panel updated media and analysts on the initiative’s progress – in the same room where they had briefed them on the original endeavor four years ago.
The highlight of this particular announcement was made public by CEO Anthony Neri, in his first keynote as leader of HPE, where he explicitly tied together the MDC strategy with HPE’s overarching emphasis now on what they term the Intelligent Edge.
“We have lapped the competition with the very best in high memory computing,” Neri said. “Memory-Driven Computing is a scalable, composable architecture which drives all our future road maps, along every product line in our business. Memory Driven Computing will be a game changer. You will be able to power your entire infrastructure with a single architecture. You will not need databases, or database licenses. And we will deliver this as a service, so u don’t have to manage it.”
These applications are still in development, but the first part of HPE’s MDC initiative is now available to customers.
“You can get started with it today,” Neri told the audience. “We are launching a Memory Driven Computing sandbox for developers. We have systems already up and running with APIs your developers need in a Linux environment. It will run on our [HPC] Superdome Flex. You will be able to compose memory, as well as compute, storage and networking.
The idea behind the sandbox is to ramp up the MDC initiative, and turn it from what the company acknowledged has been basically a science project in HP Labs to something that customers can work with and comment on, in order to accelerate the development process further.
“We announced The Machine four years ago,” said Kirk Bresniker, Fellow and VP, and, as in 2014, Chief Architect at the Hewlett Packard Labs, although he has long abandoned his clean-shaven visage then for a ZZ Top beard. “The goal was, as we looked at the changing landscape, and the preponderance of data growing at the edge – how are we going to scale that engagement and get more people into the conversation.”
That’s where introducing MDC technology to HPE Pointnext, which delivers HPE technology on a flexible consumption model, will come in.
“That’s how we scale this,” said Balaji Subramaniam Venkatesan, Director Data and Emerging Technology, HPE PointNext. “We are incubating a practice. Pointnext is taking the memory-driven concept from the Labs to the customer, so customers will get real benefits.”
The other part of the news is the announcement of MDC access through the sandbox capability on the HPE Superdome Flex, for customers and partners to be able to work with the technology.
“We have to get access to ISVs, customers and partners so we can take this even further, and we do that with the sandbox,” said Jeff Kyle, VP and GM, HPE Mission Critical Servers. “It’s scalable to get everybody started in that data driven memory driven environment, and we are happy to start it on Superdome Flex.” Given that the in-memory Superdome Flex is the most powerful server that HPE sells, that will limit the process of democratization somewhat, although its vast processing power is well-suited for the requirements of MDC.
“We have been working really hard on Superdome Flex for the big memory machine, and its sandbox is up to 96 TB of memory – larger than anything on the planet,” said Mark Potter, HPE’s Chief Technology Officer, and the Director of Hewlett Packard Labs. “Developers for years have written code assuming that memory would be scarce. What if we turned that on its head? It’s this thinking differently that is the difference. It’s not just applying the same algorithms to a larger memory pool.
“We will continue to expand the sandbox,” Potter added. “That’s why Pointnext is so important.”
The panel also included HPE’s first commercial customer for MDC, Travelport, a commerce platform that provides distribution, technology, payment and other solutions for the $7 trillion global travel industry.
“We had conversations with HPE in two places,” said Matt Minetola, EVP Technology and Global CIO, Travelport. “The first was with technology folks, to see that our problem wasn’t unique, and then we worked with organizations across HPE to seek alternatives to solve it. It’s all about solving a shopping experience – taking these concepts and doing analytics all in one integrated environment, so that you can make any adjustments for the consumer within a second. And you have to be fast enough to get on the first page of search results, because no one goes beyond that.”
In April 2018, Travelport installed a Superdome Flex in-memory computing system, and worked with HPE to rearchitect key Travelport algorithms using MDC programming techniques. HPE is also assisting Travelport to identify a performance baseline for infrastructure upgrades, drive cost-benefit analyses for the transformation journey and port, tune, re-architect and refactor MDC applications.
“Before Travelport, we had done all this kind of work with research institutes, and while these were very significant projects, they were still basically toys,” said Dr Sharad Singhair, Director, Machine Software and Applications, at HP Labs. “Here we were looking at Travelpoint’s core business applications, and looking at how to transform them. That is not a toy. This is about transforming a business application, which is very exciting. What we think we can do is bring all the data from the back end into memory. Then, as these searches become faster, a lot of these things can be turned into predictive analytics. That will allow Matt both to reduce the size of his infrastructure, and know the things customers care about when they make their reservations.”
