Advanced Computing in the Age of AI | Tuesday, August 16, 2022

HPE to OEM SGI UV 8-Socket Technology for In-Memory Processing of Linux Workloads 


An 8-socket server with in-memory of 6TB for running enterprise decision support Linux supply chain, ERP and CRP workloads in real time or near real-time: that’s the strategy behind a new partnership in which Hewlett Packard Enterprise will OEM SGI’s UV in-memory server technology.

In so doing, HPE fleshes out its “mission critical” line of enterprise compute platforms, filling the gap between its HPE ProLiant DL580 Gen9 Server, with 4-socket scalability at the low end, and the HPE Integrity Superdome X server that scales up to 16 sockets and 24 TB of memory at the high end.

The mid-range server, called the HPE Integrity MC990 X, features Intel Xeon E7-8800 v3 processors with a maximum of 144 cores and is priced starting at approximately $70,000. It’s intended for organizations looking to scale large, data-centric Linux workloads on SAP HANA, Oracle and other databases. HPE said the system is targeting enterprises that run analytics off of their CRM or ERP data, reports that normally take hours or overnight, Jeff Kyle, HPE’s director of product management, enterprise servers, told EnterpriseTech, “depending on how much data you want to crunch. But feeding that data into an in-memory analytics database solution dramatically speeds up access to reports and analysis to make better, faster business decisions in seconds rather than hours. That’s what we’re finding to be the big benefit of in-memory analytics.”

HPE's Jeff Kyle

HPE's Jeff Kyle

“We’re accelerating any Linux workload that needs to scale from 2 to 6 TB,” Kyle said. “That’s kind of a sweet spot for many enterprise customers. Not every customer is going to hit 24 to 48 TB of memory over the next year, but we see almost everyone in the enterprise space, and even some mid-sized customers, deploying in-memory solutions of between 2 and 6 TB. That’s the sweet spot for the 990.”

The in-memory marriage of HPE and SGI combines SGI’s experience building large memory footprint platforms – saving HPE the time, trouble and expense of developing an 8-socket platform on its own – while providing SGI with HPE’s global route-to-market and services reach. One industry observer told EnterpriseTech it’s “a win-win.”

In-memory computing is playing an increasing role for organizations that want real-time, or near real-time, decision making capabilities for high-volume analytics workloads, such IoT applications in which large amounts of sensor-produced data must be rapidly processed. Kyle described the Integrity MC990 X as a “bleed-over” between HPC and enterprise IT, bringing a big-memory computing platform to the more common types of data analysis solutions used in the workplace.

“‘Scale up’ used to mean the number of processors, processing power, and so forth,” said Kyle. “But for last three or four years, scale up means memory – how high can you scale memory so you can run that business application or that data management workload in memory.”

Looking forward, Kyle said, HPE is putting increasing emphasis on the ability to address the scale-up workloads. For example, researchers in Hewlett Packard Labs, the company’s R&D organization, have been working on scale-up versions of Spark, released into open source, that enable large graphics workloads to be processed completely in memory. “So instead of hundreds of small nodes going back and forth to disc and clogging network traffic, for certain Spark workloads we can move all of that into memory and have a 15X performance improvement.

The MC990 X competes with other 8 socket servers on the market from Lenovo, Fujitsu, Hitachi – and SGI.

Add a Comment