High performance computing has yet to fully embrace the cloud, but the benefits of the pay-as-you-go model is drawing in more HPC users with each passing year. One segment that has been somewhat underrepresented is engineering, which increasingly relies on HPC-powered digital simulations for product design and development. A recent survey of engineers suggests that cloud computing is now poised for more rapid adoption in this application domain.
The survey in question was conducted in the spring of 2019 by Peerless Research Group, on behalf of Digital Engineering magazine and ANSYS. As a prominent supplier of engineering simulation software, ANSYS certainly has a dog in this fight. And since the company also provides its own SaaS cloud solution, known as ANSYS Cloud, it’s particularly interested in how these users perceive the cloud delivery model and how they plan to use such services in the future.
According Wim Slagter, director for high performance computing and cloud alliances at ANSYS, the 600 or so survey respondents, all of whom worked for companies using engineering simulations, were split more of less evenly between Digital Engineering’s readership and his company’s customer base. Three company sizes were represented, with 37 percent of respondents coming from companies with 99 or fewer employees, 24 percent from those with 100 to 999 employees, and the remaining 39 percent from companies of 1,000 or more employees.
According to the survey results, only 17 percent of those surveyed are using the cloud for engineering simulation today, but an additional 20 percent are planning to use it over the next 12 months. If true and assuming that number extrapolates to the larger community, that would more than double cloud usage for engineering simulations over the next year.
For those already using the cloud on these workloads, 63 percent were using an on-premise private cloud, which depending on how respondents interpreted that category, may be nothing more than a shared in-house cluster. The remainder were split almost evenly between those using the public cloud (35 percent) and those using an ISV-managed SaaS solution, like the ANSYS Cloud (32 percent).
According to the research, the mix of cloud platforms is going to change significantly over the next year. During this timeframe, 62 percent of cloud users said they expect to be using ISV-managed clouds, compared to 45 percent that expects to use public clouds, 44 percent using a combination of on-premise with a partner-managed clouds using a regional datacenter, and 40 percent using a combination of a on-premise and partner-managed cloud using a public datacenter. The chart below illustrates the breakdown in greater detail.
Slagter says that growth appears to be backed up by what the company is experiencing with their own ANSYS Cloud offering. According to him, they are seeing what appears to be an inflection point, where certain types of customers are migrating their applications from on-premise to these ISV-based SaaS solutions.
The drivers behind this shift and the uptick in cloud adoption for engineering simulations, more broadly, can be teased out from the respondent’s answers to how they justified cloud computing to other stakeholders. These rationales varied, but the top justification, reflected by 49 percent of respondents, was that the cloud offered the ability to scale up and down quickly. That’s hardly surprising, inasmuch as computational elasticity is recognized as one of its biggest advantages.
The survey noted that over a third of users reached peak capacity with their on-premise systems a few times per month, implying that cloud resources can be used to make up for the shortfall at these times. From Slagter’s perspective, the ability to burst into the cloud is one of its most important attributes. And for engineering teams, in particular, having this capability on tap can be critical to meeting product development deadlines.
Perhaps a more interesting set of justifications is around cost. Specifically, 40 percent of those surveyed reported that lower overall costs was a significant rationale to use the cloud, while 33 percent pointed to lower maintenance costs (40 percent for those in small companies). Juxtaposed with those numbers is that cost was also cited as an important obstacle among those who are not using or planning to use the cloud for their simulation workloads.
As the study’s write-up pointed out, the cost/benefit of migrating these workloads off-premise is not a simple analysis and varies by company and application. “Not all cloud computing users see lower direct costs,” write the authors, “but when maintenance, time savings and lower capital expenditures are factored in, a compelling argument for lower costs and higher productivity can often be made.”
Time savings, in particular, was a central issue to those surveyed. In fact, increasing the speed of product development turned out to be the most important factor for evaluating engineering simulation software, in general, being cited by 81 percent of respondents. Improving engineering productivity, at 75 percent, and accelerating innovation, at 67 percent, also played into the time savings theme.
The study also pointed out significant obstacles for those considering or implementing cloud-based solutions. Security was cited as the top concern, with 59 percent saying they were worried about protecting their data and intellectual property. To some extent, cloud security has become something of a red herring, given that clouds are generally more secure than on-premise setups nowadays. Survey respondents reported they were addressing security with a variety of approaches, including evaluating security options with various cloud providers and understanding their own risk profile.
Only slightly less worrisome to these users is where their data is stored and cloud performance concerns. Each was cited by 51 percent of respondents as significant obstacles. Like security, such issues are addressed differently by different providers, so once again, the user needs to do their homework before jumping in.
The complete study may be downloaded here.