In a sign of the times, another prominent HPCer has made a move to a hyperscaler. Longtime Intel executive Bill Magro joined Google as chief technologist for high-performance computing, a newly created position that is a statement of Google’s commitment to HPC and to eclipsing the cloud competition. Magro will focus on driving HPC strategy, customer traction and success in the market.
The banner cloud companies are in a race to inject HPC-class technologies into their stacks, propelled by the demands of computationally and data-intense workloads, like machine learning and analytics.
“HPC is a major focus area for Google Cloud,” Magro told HPCwire by email. “Organizations are increasingly looking to the cloud as a means to augment existing HPC infrastructures, quickly ramp HPC capabilities for new ventures, and access the latest technologies. From universities like Clemson University and Harvard Medical School to leading companies such as Schrödinger, many organizations are already turning to Google Cloud for its usability- and compatibility-focused HPC capabilities.”
The pace of innovation at the major cloud companies is never ceasing. Google recently introduced its fourth-generation Tensor Processor Unit and is the only cloud provider to host 16-GPU A100 instances. Amazon claims the custom-developed Graviton Arm chip, FSx for Lustre, and the latest A100 GPU technologies in its cloud. Microsoft Azure offers 200 Gbps InfiniBand, the Graphcore accelerator chip, and collocation with HPE Cray supercomputers.
High-performance cloud is one of the fastest growing segments of the HPC market. HPC analyst Addison Snell, CEO of Intersect360 Research, reports continued double-digit growth and predicts even higher surges in 2020, as cloud fills in for pauses in capital expenditures caused by the COVID-19 pandemic.
Hyperscale’s rapid accretion has created a gravity that’s pulling in seasoned professionals from more traditional system and chip companies. High-performance networking innovator Steve Scott left his position at HPE for Azure in June. Azure execs Nidhi Chappell and Merrie Williamson joined from Intel last year. Barry Bolding, formerly with Cray, is now GTM leader for HPC, autonomous and quantum computing at Amazon Web Services.
Prior to his two-decade tenure at Intel, Magro worked as a physicist, running large-scale simulations on some of the nation’s largest supercomputers. “These systems — from companies like Thinking Machines, Cray, SGI, IBM, Digital Equipment Corp., and Hewlett Packard — exposed me to a wide range of architectures and application environments,” said Magro. “At Intel, I had the privilege to work with experts across the range of HPC technologies and with customers across a wide range of enterprises and disciplines.”
He sees his role as translating customer requirements into product roadmaps and strategies. “I want to help shape the evolution of the cloud to be a great place to run existing and emerging HPC workloads,” he said.
Google is one of a handful of leading cloud providers, but trails Amazon and Microsoft, in first and second place respectively. According to a report published in December, Google is gunning to be the top one or two cloud provider by 2023. There was even talk of this being an all-or-nothing approach, which Google disputed. Oracle Cloud, IBM Cloud, Alibaba and Tencent are also jockeying for position in this competitive space. For the last two years, Google Cloud has been led by former Oracle executive Thomas Kurian.
With the addition of Magro to its executive team, Google now has a group leader for HPC systems. His technical and business expertise and connections in the industry put him in position to change the playing field quickly, said one market-watcher we spoke with.