With roughly 75% of American companies stating that high performance computing is critical to competitive success, why is it that so many end users have a difficult time justifying their investments in HPC?
The arguments in favor of increasing the amount spent on high performance computing can be quite numerous, depending on industry. However, while some at the high end of the market in terms of traditional investment in supercomputing (oil and gas, financial services, national lab-level research) might have an easier time given the historic ROI chain and clear competitive advantages of computational speed and capacity, the entry level and to a slightly lesser extent, mid-range, face an uphill battle.
When it comes to how companies justify their investments, Intersect360 found in a recent Council on Competitiveness survey of roughly 160 American firms in multiple market segments that 24 percent were able to demonstrate a time-to-market value. 23 percent said that their argument was that without HPC they would not be able to solve the problems at hand.
While other reasons, including utilization rates, improvements in quality, features, and capabilities, and reduced costs compared to physical methods (i.e. actual crash tests versus simulations) were common, the time to market remained the leader—although not a runaway lead in the investment justification department.
The graphic on the left highlights the breakdown Intersect360 found among respondents, which highlights both the diversity of justification arguments and the fact that there is no real “bullet proof” HPC investment strategy that holds up across market segments.
As you can see in the graphic below, when it comes to finding funds for HPC investments and justifying increasing investments, companies overwhelmingly are finding some real difficulty securing the resources they need to drive competitiveness. Again, while this is not necessarily a surprise given the constraints of larger budgets, this group of entry-level HPC users represent a broad swath of small businesses that make up part of the core of American firms that are nimble enough to drive new innovations, add jobs, growth, and new ideas or products.
In these cases, size does not necessarily make as much difference as one might think. Even for the large supercomputing centers, there can be some pushback or resistance and of course, at smaller companies where budgets are tighter, securing buy-in can be far more difficult. The disparity between organizational size in the graphic below and the above chart is interesting to note, especially in terms of how the mid-range level perceives their justification. The lightest blue represents the most ease with the darkest showing the most challenging in terms of securing investment dollars required for bolstering HPC capability.
In their responses, participants were able to offer longer explanation about certain issues, including the justification problem. As one user in manufacturing noted, “If I look at the top ten supercomputers in the world, they’re not entirely out of reach of certain corporations. But if I get this machine, will we really see the benefits in our design process? And that has become a chicken-and-egg problem because management can be skeptical, and of course they want a justification for such a large investment.”
As Intersect360 finds, however, HPC experience is important to questions like the one above. With the software complexity of scaling to new levels of compute, not to mention the many other investments outside of pure hardware and people, it’s also a matter of power, licensing, and other costs. Although the more experienced centers and companies typically have a process for examining these issues in some detail, smaller firms appear to be having a harder time pulling together all the pieces to make an argument for HPC.
But at the higher end, as another participant in the survey from the aerospace industry noted, “You cannot come to me and say, ‘Here’s a problem I want to do and I don’t have the computational resources to do it,’ because we will go get them. If that means substantially more cores in the company, if that means buying a bigger computer, if that means buying additional software, the return on this investment is easy for us to justify.”
At the end of the day, it comes down to a matter of competitiveness, but how can smaller companies demonstrate that value if the full force of HPC cannot be justified enough to be put into full practice? We heard quite a bit about this during an engineering and simulation panel at EnterpriseHPC (more on that here) but even there, as with this survey, the breakdown of both justification and implementation was split across multiple areas—people, ISVs, scalability of software, TCO—meaning that there really is no bullet proof way to approach companies with HPC justification arguments.
Nonetheless, the stories of how HPC investments have played out over the course of company lifecycles and bolstered competitiveness might be the secret key—or at least a start at communicating the commercial value of high performance computing.