Reinventing HPC


Nages Sieslack: After two and half years since the Cray acquisition, how has this addition impacted your HPC business?

Bill Mannel: Acquiring Cray has greatly benefitted our HPC and AI business. The combination of the Cray EX product line with the HPE Apollo product line created the premier HPC and AI portfolio in the industry. The Cray acquisition also allowed us to provide a complete solution for customers, including our HPE Slingshot fabric, ClusterStor HPC storage, and the Cray HPC Software stack including Cray System Management and the Cray Programming Environment. This has allowed us to provide solutions to customers that each individual company alone could not. We can tailor our HPC and AI solutions to fit the customers’ needs and requirements. The benefits go beyond products as the combination has brought together the leading minds and experts in the HPC and AI industry all under the HPE banner, including the largest applications and performance team.

Sieslack: Can you tell us anything about your plans for the Slingshot interconnect that was acquired as part of that acquisition and how that technology will fit into your broader HPC and AI portfolio?

Mannel: We are expanding the power of the Slingshot interconnect to include both Cray supercomputers and volume HPC clusters. It turns out that solving problems of performance and efficiency at large scale means you also solve these problems for smaller scale systems, too.

We now support:

HPE Apollo and HPE ProLiant servers (not just HPE Cray EX systems); Commodity cluster operating systems, such as RHEL, SuSE, and Rocky Linux.

Whether a volume cluster or a supercomputer, HPE focuses on delivering a complete cluster solution backed by excellent support.

We now also offer the Cray Programming Environment for volume clusters. Cray PE will take full advantage of Slingshot capabilities, and because it is MPICH ABI compatible, Slingshot can run applications written for other MPICH derivatives.  (Note that we expose our unique functionality using Libfabric APIs—a standard of the OFA—the broader open-source community can innovate around our capabilities as well.)

Finally, we add Cray System Management as a management solution for Apollo and Proliant HPC solutions, allowing a microservices-based cluster manager for efficiency in managing large to very large clusters.

Sieslack: What unique attributes does HPE GreenLake for HPC bring to cloud services and what has been its initial reception from customers?

Mannel: HPE GreenLake for HPC offers a unique model for bringing the cloud experience to on-premise or colocation users, allowing customers to relate to the infrastructure as a turnkey private HPC cloud service.

Customers get all the attributes of a primary cloud service. It’s consumption-based and can be fully managed and operated by HPE. But in addition, they also get key HPC building blocks (e.g., parallel file system, low latency interconnect, GPUs).

From manufacturing to bioscience, everyone today wants—and needs—to consume HPC as a service. And HPE GreenLake for HPC gives them that capability. Even among smaller institutions and research universities, GreenLake is a key tool for sharing computing capacity across different faculties or organizations.

We also notice that most customers require the integration of public cloud services into their HPC environment. The recent GreenLake for HPC release offers hybrid cloud and multi-cloud services support. We are thinking of growing the ecosystem around the HPE GreenLake for HPC cloud service by enabling partner and customer software and services integrations by introducing HPE GreenLake services APIs with multi-cloud capabilities.

Sieslack: What are your expectations of the growth trajectory of HPE GreenLake?

Mannel: With the current attach rate of requiring services for HPC opportunities and HPE GreenLake for HPC incorporating flagship HPC technology building blocks, we believe opportunities for growth will skyrocket in the second half of 2022 and 2023. We see many RFIs/RFPs for 2023-2024 where HPE GreenLake would be a leading response. Any enterprise, regardless of size, who wants their HPC environment and converged HPC/AI environment to be consumed as a service will find HPE GreenLake for HPC providing a best-of-two-worlds offering.

Sieslack: Which emerging technologies or architectures are most likely to impact HPC in the next few years?

Mannel: The next few years will bring an explosion of new offerings and choice to the HPC and AI industry. We believe the upcoming years will see increased merging and mixing of traditional HPC and AI workloads and less siloed applications. This will bring to market new hybrid solutions and architectures that can handle both types of workloads seamlessly without sacrificing performance. Many environments will be increasing hybrid, combining on-premises resources with cloud environments. But many customers will demand more than is provided in a single public cloud offering, driving demand for multi-cloud solutions. The lines between a CPU and accelerator will continue to blur to accommodate this new working model. Our solutions and offerings will continue to lead this transition, providing performance and availability unmatched in the industry.

Visit the HPE Team at Booth E509 to speak to their team.

Go back