Dell pushing hybrid quantum/classical system in HPC overhaul

Steady now. It's not the full-fledged quantum that will suddenly let you crack encryption algos

SC22 Dell is expanding its high performance computing (HPC) portfolio with the addition of a quantum computing platform - where customers can start testing out quantum algorithms - plus an overhauled APEX service and PowerEdge systems based on Intel's Sapphire Rapids next-gen processors.

Unveiled to coincide with the SC22 conference in Dallas, Dell said its updated technologies and services are to help customers power demanding applications. One intriguing part of the announcement is the Dell Quantum Computing Solution, which is a hybrid quantum/classical platform delivered using the company's PowerEdge servers combined with quantum technology from IonQ.

Dell's pitch is that it is a platform to introduce biz customers to the quantum world and not a full-fledged quantum system which is going to suddenly enable you start cracking encryption algorithms.

"We've done a great deal of finalization of the solution offering as well as the integration with IonQ in order to make it a full product that customers can now buy, in what we consider the journey towards quantum, which is about putting hands on keyboard and learning, performing experimentation,” said Ken Durazzo, vice president of Dell’s CTO Research Office.

"So you drill into the learning and can start to find out which applications in your environment are more likely to leverage quantum acceleration, and then you'll build proof of concepts, and finally productize," he added.

The Quantum Computing Solution comprises two parts. The first is the Dell hardware and software environment including a quantum simulator, made up of one or more PowerEdge 750xa server nodes, which can be deployed at the organization’s premises. The second is access to IonQ's Aria quantum computer via the internet, for when customers are ready to test on actual quantum hardware.

"The customer will be able to perform any experimental simulation here on-premise[s], and only when they need to access an actual physical quantum platform will the traffic actually go over the internet. So this part is fully contained, including simulators that allow a customer to learn and experiment locally. And when they want to see how the application will be accelerated on the real platform, then they would run it across the internet," Durazzo said.

The Qiskit Dell Runtime lets users build quantum applications in a "write once, run anywhere" fashion so that programming and the way that code executes is the same regardless of where it runs on the virtual or physical quantum processor.

Availability of the Dell Quantum Computing Solution will be the US and Canada first, but rolling it out to other territories over the course of the next year.

Meanwhile, Dell is expanding the HPC systems it is making available via its APEX as-a-service portfolio. The new Dell APEX High Performance Computing is now a much more comprehensive suite, according to the company’s Director of HPC Product Management, Armando Acosta.

"When you looked at the original offer, it was just essentially, hey, we can do HPC, we give you some compute, we give you some storage. What we're doing here is we're essentially giving you the full managed stack with HPC," he told us.

The way it works is like a building block approach, Acosta said, using validated designs for financial services, life sciences or manufacturing, for example.

"We created a validated design where we tell you, here's your compute building block, here's your network building block, here's your storage building block and then here's how we optimised it, based on running a benchmark. And now we can tell you essentially, based on some performance guidance on your side, or capacity, here's what we can build for you," he told us.

Rival HPE began selling supercomputing-as-a-service via its Greenlake platform several years ago.

Organizations can choose from several compute options from Dell, so computational fluid dynamics might call for memory intensive needs, Acosta said, and customers can choose to add GPUs for accelerated workloads. The storage options include the choice of a shared NFS running on PowerEdge servers or Dell PowerScale NAS, while connectivity starts with 100Gbps Ethernet with InfiniBand an option.

Using the Dell Cloud Managed Services Console, customers can choose what type of jobs they want to run and incident management, and this integrates with Bright Cluster Manager for monitoring and alerts.

Acosta said that Dell is working to integrate its Omnia HPC stack into this APEX service. This is basically Dell’s build of the OpenHPC open source toolkit, which uses Ansible playbooks to deploy environments using Kubernetes and Docker, or a more traditional HPC environment using the Slurm Job Scheduler.

“So if you just have your traditional users that want to do modelling and simulation, we can give you a traditional Slurm environment. But if you're a data scientist, or somebody that's doing data analytics, then we set up a Kubernetes and Docker environment for you in that aspect,” Acosta explained.

For now, Dell APEX High Performance Computing is available only in the US, with broader availability coming later.

Also at SC22, Dell announced new PowerEdge servers. The PowerEdge XE9680 is a dual-socket system that will feature a pair of Intel’s 4th Gen Xeon Scalable processors plus eight Nvidia H100 or A100 Tensor Core GPUs.

The PowerEdge XE9640 pairs Intel Xeon processors with four GPU Intel Data Center GPU Max Series accelerators, and the PowerEdge XE8640 features two 4th Gen Xeon Scalable processors and four H100 GPUs.

These new PowerEdge servers have planned global availability for sometime in the first half of 2023, in line with the 4th Gen Xeon Scalable chips themselves. ®