Many companies and research organizations have a powerful desire for their researchers’ impact not to be constrained by the size or type of computing environment. One clear answer to this is cloud computing, which frees the researchers’ innovation with access to agility and scale.
Users of HPC and AI often deliver some of the greatest value, however these demanding users of computing are hungry for better services and capabilities than is offered by traditional cloud platforms. In response, Microsoft has built the next generation HPC-first cloud, meaning a cloud platform designed purposefully for HPC and AI workloads. This brings together supercomputer performance, familiar infrastructure features, and customer interactions expected by the HPC community.
In this new Microsoft Azure cloud, HPC and AI workloads are the plan, not plugin or afterthought. To achieve this feat, Microsoft is assembling an HPC dream team with several industry leaders, led by Nidhi Chappell, Head of the Azure HPC & AI product and engineering team at Microsoft, and formerly head of Intel’s work on AI and HPC in the cloud. Nidhi’s team includes HPC luminary Andrew Jones and HPCwire’s 2020 Person to Watch Evan Burness. Every member of the team has hands-on experience with HPC workloads, and the team is growing.
“From the start, we focused on creating the best infrastructure for HPC workloads to create the industry’s first HPC optimized cloud – a cloud designed for HPC workloads and for HPC users” said Chappell.
“We are replicating in the cloud what our clients are using in on-premises clusters,” she added. “It’s all about delivering genuine HPC technical capabilities for real world applications.”
Microsoft’s HPC team is very familiar with how to ensure HPC workloads and users can achieve their potential. One example of a major differentiator, Chappell says, is that Microsoft’s new HPC cloud incorporates typical HPC technologies and processes into the Azure cloud service at scale.
“Traditional clouds promising HPC often require users to fit their processes to the cloud, rather than enable customers to use HPC in the same way they do in their on-premises data centers,” she said.
The great potential of the Azure HPC cloud is to combine that familiar baseline with the new possibilities opened up by the cloud, such as technology agility, workload scale, geographical resilience, operational freedoms, and the potential for cost-savings.
However, Microsoft recognized that customers’ needs went beyond technical requirements to encompass business needs, such as improving operational processes and planning cycles. Azure HPC supports streamlined processes and tools to help customers with planning for HPC in the cloud, such as cost models, proofs of concept, identification of value opportunities and working with refresh and procurement processes.
Microsoft’s new HPC Cloud empowers HPC and AI practitioners to exploit the latest and greatest hardware for shorter time to insight, at scale to allow more simulations in parallel for better quality of decisions, along with the familiarity to enable low risk adoption of cloud.
“We are bringing the best technologies and people to the table,” said Chappell “so that Azure HPC cloud can free our customers to achieve innovations they were previously only able to dream of.”