How to accelerate innovation and insight by defeating the HPC "storage monster"

[UliPlechschmidt]

HPC storage is often the culprit when compute systems can't deliver. Cray ClusterStor E1000 storage is purpose-built to address those challenges—and it's now available in enterprise data center racks.

In the digital economy innovation and insight need to happen at supercomputing speeds. And when “systems of innovation and insight” can't deliver, HPC storage is often the culprit.  

Here are some of the reasons why:

• Keeps expensive CPU- and GPU-accelerated compute nodes sitting idle waiting for data that can't be served fast enough for processing.
• Consumes an ever-increasing percentage of the overall systems budget—at the expense of CPU- and GPU-accelerated compute infrastructure
• Slows time-to-problem identification and resolution if the storage and compute infrastructures are delivered by different vendors

This past July, HPE introduced the Cray ClusterStor E1000 storage system as an HPE product. The ClusterStor E1000 was purpose-engineered to solve all those challenges listed above. And since its introduction, HPE has shipped or signed more than 2 exabytes of ClusterStor E1000 storage capacity.   

ClusterStor Fun Facts

Three new HPE Cray EX supercomputers with Cray ClusterStor E1000 storage—Frontier, El Capitan, and, LUMI—will have more combined computing power than all of today's top 500 fastest supercomputers put together

5 out of 7 global oil supermajors use Cray ClusterStor systems today with 4 already investing in the new Cray Clusterstor E1000 storage system

For more on HPC storage challenges and how the Cray ClusterStor E1000 storage system uniquely solves them, this video offers a quick rundown on everything you need to know.   

HPC storage for enterprise data centers

Many enterprises who are using HPC based on HPE Apollo CPU or GPU-accelerated systems have been waiting for Cray ClusterStor E1000 supercomputing storage technology to become available for enterprise data center racks. The wait is over!

Originally launched in custom storage racks, the ClusterStor supercomputing storage technology is now available in enterprise data center racks. You can build a Cray ClusterStor E1000 storage system with the architecture shown in the following graphic in less than half of an enterprise data center rack. From there, scale to wherever you need to go—without architectural limitations.

Build a Cray ClusterStor E1000 storage system in less than half of an enterprise data center rack

Independent analyst firm Hyperion Research thinks that this approach can serve as a blueprint for architecting an HPC storage strategy to provide optimal performance for a wide range of diverse workloads.*

Now enterprises of all sizes can use that technology in enterprise data center racks to drive accelerated innovation and insights. Both of which need to happen at supercomputing speeds in today's digital economy.

For more information on how the ClusterStor E1000 can help you:

• View this infographic on the Cray ClusterStor E1000 storage system. 
• Read the business white paper "The New HPC Era Needs New HPC Storage."
• Take a deeper dive into ClusterStor E1000 storage technology in this technical white paper.

 * Hyperion Research, Quick Take: Effectively Managing HPC/HPDA/AI Storage Requirements: An Increasingly Critical Challenge, November 2020

Uli Plechschmidt
Hewlett Packard Enterprise

twitter.com/hpe_hpc
linkedin.com/showcase/hpe-ai/
hpe.com/us/en/solutions/hpc