HPE Discover just took place in Las Vegas. President and CEO Antonio Neri started the proceedings with his sixth keynote as the company’s leader, reflecting on past strategic moves. HPE bet the farm on edge computing, tripled its networking business with the acquisition of Aruba, acquired Cray to accelerate AI and was the first to offer everything-as-a-service (XaaS) with GreenLake. In fact, GreenLake was the main act at this event. Read on as I sort through the various announcements and initiatives from the show.

You can use HPE’s supercomputers as a service for AI

HPE announced HPE GreenLake for Large Language Models; the cloud service for LLMs provides access to a Cray XD supercomputer, Nvidia H100 GPUs, a full software stack and services enabling customers to run machine-learning applications remotely. In the same way that GreenLake packages computing as a utility service and charges for usage, HPE will charge for AI compute consumed in the cloud.

This initial service for LLMs appears to be part of a larger vision to introduce AI services for domain- and industry-specific applications in climate modeling, healthcare and life sciences, financial services, manufacturing and transportation. GreenLake services can be run entirely on customer premises or in a hybrid on-premises and cloud environment. These new services, still under the GreenLake brand, will be hosted fully in the cloud Canada for North America. The LLM model is provided through a partnership with German AI startup Aleph Alpha, whose Luminous software is pre-trained and optimized for analyzing and processing data at scale. I am told that customers can bring their own models and am expecting announcements further down the line with companies like Hugging Face.

I still have many questions on this private cloud based LLMs. Current questions include:

• How are these XaaS “instances” being purchased? By the hour? By the resource like a GPU? How are they priced?
• When will these be GA in the US? WE? Asia?
• Will QScale be serving all of NA or just Canada?
• Which Cray systems are these sitting on top of?

Infrastructure and software to accelerate AI training

Continuing with the AI theme, HPE is now offering its HPE Machine Learning Development Environment on HPE GreenLake for HPC, providing optimized AI infrastructure and software to accelerate AI training. HPE has recently collapsed four HPE Ezmeral product lines into two—HPE Ezmeral Software data fabric and a new Unified Analytics product. This makes it easier and more cost-effective for customers to install, build and deploy AI and ML applications.

Automating IT services management

OpsRamp, acquired by HPE in May 2023, is an AI tool used by managed services providers to automate IT services management. It is hybrid multi-cloud ready, supporting AWS, Microsoft Azure and Google Cloud along with infrastructure from vendors including Dell Technologies, Cisco Systems, NetApp and Pure Storage.

HPE is now integrating OpsRamp as a SaaS offering directly into HPE GreenLake with full-stack observability. OpsRamp extends HPE’s observability capabilities beyond the HPE infrastructure across a multi-vendor, multi-cloud infrastructure all the way up to the application.

Hybrid and multi-cloud environments present a lot of complexities across assets, from on-premises to the public cloud. I like it that HPE is making moves to simplify hybrid complexity and drive revenue wherever the compute is.

A click-to-deploy virtual machine private cloud service

GreenLake for Private Cloud Business Edition (PCBE) is a new SaaS version of the GreenLake private cloud. PCBE is based on HPE’s disaggregated hyper-converged infrastructure solution (dHCI). HPE aims to deliver the hardware that powers the private cloud service to customers within 12 business days of ordering, unlike the fully managed GreenLake for Private Cloud Enterprise, which comes with terms and conditions that can take months to sort through.

Pre-provisioned GreenLake services at Equinix

HPE also announced it is expanding its partnership with Equinix to provide GreenLake pre-deployed private cloud solutions to address a broad swath of private cloud requirements. Equinix is a global colocation data center provider with seven supported HPE locationss in San Jose, Washington, D.C., Toronto, London, Frankfurt, Sydney and Singapore.

While HPE wouldn’t share specifics on how many weeks or months of pre-provisioned equipment is available, it appears the goal to get customers up and running should be measured in terms of days versus of months.

Next to the LLM as a service announcement, the Equinix expansion was my favorite, if nothing else, to show just how far HPE has taken the hybrid cloud. It should get IT thinking about the real difference between the public and hybrid cloud.

Making HPE-AWS hybrid, multicloud deployments easier

New releases from HPE should make HPE-AWS hybrid cloud deployments easier by adding consulting services, application development, modernization, migration and managed services. New offerings in the AWS Marketplace are the HPE Nonstop development environment and HPE Fraud Risk Management.

Also new is the ability to self-manage virtual machines deployed in AWS through HPE GreenLake for Private Cloud Business Edition, container support for Amazon EKS in HPE GreenLake Private Cloud Enterprise and

backup capabilities for the Amazon RDS database through HPE GreenLake for Backup and Recovery.

Wrapping up

In 2019, HPE announced a strategy to deliver everything as a service through HPE GreenLake. Since then, that has proven to be a winning strategy as the hybrid cloud has gained momentum. During his keynote, Antonio Neri showed some impressive numbers and statistics demonstrating that this is now a real business. I like that HPE continues to support the hybrid cloud with a steady stream of pay-per-use offerings. I would like to see more cross-cloud services that can be used on all public and hybrid clouds.

The event’s most significant and topical announcement was the introduction of HPE GreenLake for LLMs. The new cloud offering will run on Cray supercomputers, allowing enterprises to train, tune and deploy large-scale AI based on individual business data—a move in the right direction for HPE. Still, it remains how it will stack up against other cloud-based services running LLMs, such as Nvidia’s NeMo and Amazon Bedrock.