Buyers and users of HPC have decades of hard-won experience and best practices. This ranges from a community understanding of what HPC is, to what constitutes reasonable performance targets. The skill of a good HPC leader is to build on these long-established baselines whilst seeking innovations to free the creativity and value generation of the users.
When exploring different technology options, HPC managers will often think of three key aspects: the benefits (such as performance); the costs; and the risks or uncertainties.
“You can imagine different options being positioned relative to each other in this 3D space of benefits, cost and risk” said Andrew Jones, Principal Program Manager for strategic planning in the Microsoft Azure HPC & AI group.
“Change along any one axis means changes in one or both of the other axes,” he said. “For example, increasing performance leads to an increase in cost or risk, whilst reducing risk requires backing off on performance or investing in mitigations.”
Understanding customer viewpoints and experiences like this is a key feature of the Azure HPC cloud that Microsoft is building. As well as delivering genuine HPC capabilities in Azure that compare well with traditional HPC installations in performance, Microsoft’s approach of employing familiar technologies is designed to reduce the risk for buyers of HPC, whilst releasing the users’ potential.
The vision is for Azure to be one of the world’s leading providers of HPC solutions, not just one of the leading providers of cloud computing for HPC. Jones’s role is to help map out that path, from today where Azure has a range of HPC-focused offerings, to a future where Azure is as synonymous with trusted high quality HPC as any of the other leading HPC brands over the years.
“We are building teams of HPC practitioners to help customers throughout their cloud journey, from early business planning, to solution delivery, and in-service HPC support and expertise” said Jones. “We are also starting to engage with customers in the way that traditional HPC providers have done, rather than expect customers to change their business processes to fit the cloud providers”.
“We encourage any customers who can share their needs for the future of HPC cloud to get in touch” encouraged Jones.
Microsoft’s strategy can be described as a proactive, collaborative partnership with customers that begins in the early days of the HPC planning cycle.
“The HPC planning cycle is a long one, spanning the identification of relevant future technologies, establishing user requirements and innovation opportunities, securing funding, de-risking through technology evaluation, running procurement processes, and actual deployments,” said Jones, who spent over a decade of his career consulting and training HPC leaders on how to manage value and risk, to calculate total cost of ownership (TCO), to effectively plan and evaluate options, and other key aspects of the business side of HPC.
Jones says that the cloud story has moved on enormously since the often less than stellar early experiences with the various cloud providers, but there is plenty of work to do. “A key part of my decision to join Microsoft was the vision I saw at Azure to do HPC properly, and the desire to build the right teams to make it happen” he said.
There’s a rising confidence factor in the HPC community already that the Microsoft Azure HPC people truly know their stuff, with customers reporting that this shines through in recent engagements.
This all bodes well for Azure HPC to continue delivering customer advantages in all three axes: sector leading performance, competitive cost structure, and best-in-class risk reductions through an enterprise-class and familiar HPC experience.