HPE Shows Single-Memory Computing Breakthrough
The prototype system features an unprecedented 160 terabytes of memory, allowing it to simultaneously handle every piece of data from Facebook
Broad range of applications
The company said it isn’t planning to produce a single product directly out of its research, but rather to apply its findings to numerous types of systems.
“The architecture we have unveiled can be applied to every computing category – from intelligent edge devices to supercomputers,” said HPE chief technology officer and HP Labs director Mark Potter.
The prototype uses a Linux-based operating system running on ThunderX2, a second-generation dual socket-capable ARMv8-A system on a chip produced by Cavium.
Key to making its in-memory focus work are the high-performance fabric protocol, photonics/optical communications links, including a new X1 photonics module, and software programming tools designed to take advantage of large amounts of persistent memory.
Cavium, the chip maker who manufactured the system’s ARM-based chip, said the system architecture could be used for next-generation data centre, cloud and high-performance computing applications.
The announcement comes a day after HPE introduced compute, storage and networking software packages running SAP’s in-memory HANA database, which likewise takes advantage of in-memory computing to address large amounts of data.
How well do you know the cloud? Try our quiz!
