Owen Thomas, Red Oak Consulting

Interest in HPC in the cloud continues to grow, but those who can see a strategic benefit are often disappointed when a TCO for cloud vs on-premises is undertaken.  Why? Because existing TCO models utilised for comparison are not totally appropriate for HPC.  This talk will look at some of the reasons these and more traditional TCO models need refinement for a cloud solution.

We’ll also look at what contributes to the oft-quoted delta in price per system billing unit between on-premises and cloud HPC? What sets of assumptions are reasonable to make and just where does your institution lie on a range of highly variable cost-bases? TCO models are certainly helpful when comparing cloud and on-premises services but we will also explore why we think that sometimes a ‘TCO model is not enough’.

We’ll discuss why we believe it’s important to attempt to model opportunity costs in an organisation. This also includes taking an enlightened view on the value of money when making business decisions, even in an R&D context. We show that a better understanding of value verses cost, in the context of migrating HPC capability to the cloud, better exposes the value of cloud based HPC services.