Hewlett Packard Enterprise (HPE ) has unveiled what it claims is the world's largest computer with unified memory, marking a new milestone in 'The Machine' research project. The Machine is the largest research and development program in the company's history. In it, HPE is developing a new computing paradigm called Memory-Driven Computing - a computer architecture developed specifically for the age of Big Data.

Memory-Driven Computing places the memory, not the processor, at the center of the computer architecture. By avoiding inefficiencies in the interaction between memory, disk storage and processors, the new technology aims to reduce the time required to solve complex problems from days to hours, hours to minutes and minutes to seconds.

"The next scientific breakthrough is hiding behind the mountains of data we are generating," says Meg Whitman, CEO of Hewlett Packard Enterprise. "We need a computer built specifically for the age of big data."

© Hewlett Packard Enterprise

Practical example

The prototype presented has 160 terabytes of memory and should be able to process five times the data of all the books in the Library of Congress at the same time - around 160 million books. According to HPE, it has never before been possible to store and process entire data sets of this size in a system with standardized memory.

Based on the current prototype, HPE expects that the architecture can easily scale up to an exabyte-sized unified memory system, and beyond that to a nearly limitless 4,096 yottabytes of memory. By way of comparison, this is 250,000 times the current total amount of data in the world. This should allow, for example, the data of all autonomous vehicles worldwide or the data of all space research to be analyzed simultaneously.

"We believe that memory-driven computing is the solution to advancing the technology industry in a way that will enable us to make great strides in all areas of society," said Mark Potter, CTO of HPE and Director of Hewlett Packard Labs. "The architecture we've introduced can be used in any category of computing - from smart devices in the Internet of Things to supercomputers."

The technical data

The new prototype is based on the results of The Machine research project. This includes:

• 160 TB of shared memory across 40 physical compute nodes, interconnected with a high-performance fabric protocol
• An optimized Linux-based operating system running on ThunderX2, Cavium's second-generation ARMv8 SoC architecture.
• Photonic/optical communication links, including the new X1 Photonics Module, and
• Software programming tools to take full advantage of the abundance of non-volatile memory.