Cloud Simulation: Where Are We Now?

The cloud is promising to deliver high performance computing (HPC) capabilities to the masses, for very little money. Running a complex direct numerical simulation on an engine combustion event? Or maybe you just want to run your basic structural simulations, but 1,000 times faster than your desktop workstation or on-premises servers can manage. The cloud can help you out with that. In this article, we are going to take a look at the current state of cloud simulation, and we’ll be asking a bunch of industry experts to weigh in on the current state of play for the market and the technology itself.

The Cloud

But first, let’s just define the cloud a little bit. The cloud, in a nutshell, is an IT paradigm allowing users to tap into resources located in offsite data centers. Those resources can come in the form of extra computational power, via HPC clusters, or in the form of software that is hosted offsite and can be run in a user’s browser, eliminating the need for installation on a local machine. This can be an attractive proposition to companies using simulation software, as these cloud data centers can deal with all the hardware and infrastructure issues, leaving companies to carry on with what they do best (engineering) without worrying about enterprise-level software deployments, constant onsite server upgrades and other issues.Instead of the old IT paradigm, which was heavily focused on offering software as a product, the cloud paradigm instead offers its wares as a service, with users effectively paying for what they use, rather than constantly paying for licenses, CPU power or additional hardware.

Engine model made in EnSight. Model shows how fuel sprays into an internal combustion engine. (Image courtesy of ANSYS.)

The Hype Is Over (Let’s Do Some Work)

The Gartner Hype Cycle is a reasonable yardstick for measuring hype and expectations of emerging technologies, as well as giving some indication of a technology’s adoption. Brand-new technologies have low adoption, as they tend to be in the early phase of development and are generally used by research institutes and product developers. As the hype varies over the lifetime of the technology, and as the technology moves into maturity, the product receives wider adoption when bugs are ironed out and public trust in the technology increases.According to the 2017 Gartner Hype Cycle for Cloud Computing, cloud computing itself has finally emerged from the “Trough of Disillusionment” onto the “Slope of Enlightenment.” This means that, finally, after the oscillating expectations associated with emerging technologies, the hype is finally dying down, and we should now see a period of more widespread adoption of these services, before moving onto a state of increased productivity.

Moving out of the trough.

For those who are unfamiliar with how the Hype Cycle works, the Gartner website offers the following summary:

Innovation trigger: A potential technology breakthrough kicks things off. Early proof-of-concept stories and media interest trigger significant publicity. Often no usable products exist, and commercial viability is unproven.

Peak of inflated expectations: Early publicity produces a number of success stories—often accompanied by scores of failures. Some companies take action; many do not.

Trough of disillusionment: Interest wanes as experiments and implementations fail to deliver. Producers of the technology shake out or fail. Investments continue only if the surviving providers improve their products to the satisfaction of early adopters.

Slope of enlightenment: More instances of how the technology can benefit the enterprise start to crystallize and become more widely understood. Second- and third-generation products appear from technology providers. More enterprises fund pilots; conservative companies remain cautious.

Plateau of productivity: Mainstream adoption starts to take off. Criteria for assessing provider viability are more clearly defined. The technology’s broad market applicability and relevance are clearly paying off.

So cloud computing is now on the slope of enlightenment, according to the analysts. But what does industry have to say about the current state of cloud computing and simulation?

Industry Views

Engineering.com spoke to Autodesk, Ansys and Rescale, three vendors approaching the cloud from different perspectives.

Autodesk, who historically has been a design-focused company, is interested in bringing the power of cloud-based simulation into the hands of designers and has had a cloud strategy for about 5 years. Currently, Autodesk offers a range of cloud-based simulation packages, including Autodesk CFD, Moldflow and, of course, Fusion360. Autodesk provides cloud computing time to customers via use of “cloud credits,” which can be purchased to exchange for simulation or rendering time from Autodesk cloud services.

Ansys, in contrast to Autodesk, is a simulation company first, and so it’s focused on allowing engineers to assess more design variations and make better, faster engineering decisions via cloud resources. And naturally, Ansys is all about making simulation better and more robust. Ansys offers cloud computing time via use of existing licenses or by use of flexible “elastic units” as part of the Elastic Licensing model (effectively a prepay model with the option to transition to post pay if the units are used up). Elastic Licensing allows users to access most cloud-enabled software from Ansys.

Rescale is a startup founded in San Francisco in 2011, and in that short time, it has become the largest HPC infrastructure network in the world, with over 8 million servers and 1,400 PFLOPs available for public simulation usage. The Rescale ScaleX platform offers over 250 different software packages, all accessible on the cloud and with HPC computational power available on tap from the Rescale data centers. The platform allows users to pay for the hardware they use and the time they use it. They can pay for software per use or bring their own license. Rescale provides per-second billing of compute resources and is driving towards per-second billing of on-demand software licensing as well.

Challenges and Adoption

It seems that the engineering fields are late-comers to the cloud computing party.This appears to be due to a mix of cultural and technical reluctance from companies unwilling to take on new software platforms.

“Data security has been a notorious concern, particularly for larger engineering organizations,” said Todd McDevitt, director of product management at Ansys. “This is improving, but it is still a barrier for specific industries like A&D (aerospace and defense) and semiconductor.”

These industries are traditionally conservative in terms of adopting new paradigms, due to national security and intellectual property concerns. If they want to migrate their data to the cloud, they need assurance that their data is going to be safe.

But data security is not the only hurdle to mass adoption. As simulation capability grows, so do the demands that we put on the simulation software—we generate more data as our simulations become more detailed and precise.

“Large simulation output files are a challenge,” continued McDevitt. “Ideally, engineering teams adopting the cloud will choose to leave their data in the cloud and manage it there; however, this is often not the case. Most organizations will eventually want to download the simulation output files and manage them longterm on their on-premise systems. These customers are often faced with download times on par with the time required to generate the data.”

Autodesk seems to be in agreement with the above sentiment.

“One of the biggest challenges to cloud simulation is the upload and download of large datasets,” said Greg Fallon, vice president of simulation at Autodesk. “One of the challenges when doing a large, complex simulation is that after you upload the large file to the cloud and generate a gigantic data setup on solving a problem, traditionally you have to download that data in order to view it or make it usable. I think that’s ubiquitous if you talk to any major super computing center around the world. They all run into that problem. That’s one of the great technological problems.”

But adoption is picking up as the technology matures and customers become aware of the value that can be gained from using the cloud for simulation.

“Over the last few years, I’ve seen a major shift in attitudes from companies,” continued Fallon. “Two years ago, there was a lot of reluctance to use the cloud for engineering applications because of IP risks. But more and more CIOs of big corporations are starting to switch. People are starting to realize that, in many instances, the data in the cloud is just as safe, if not safer, than data that’s stored locally.”

Compute Sells … But Who’s Buying?

According to Gartner, the period during which a product moves out from the “Trough of Disillusionment” and onto the “Slope of Enlightenment” coincides with an uptake in adoption of that product.

As confirmed by the industry experts we spoke to, this is a fairly accurate representation of cloud simulation in 2018.

“Although more than 90 percent of the current market is on-premises, HPC is at an inflection point today, where many of the traditional on-premises customers are choosing cloud as a critical capability for some or all of their current and future HPC needs,” said Dave Anselmi, director of product management for Rescale.

“Historically in HPC, performance has increased about 100 times every 10 years (Moore’s law) for similar cost,” continued Anselmi. “As prices have come down, simulation complexity has also increased. Today’s simulation will run tomorrow faster and cheaper. But simulation requirements keep rising, with more complex models, more complex physics and multidisciplinary use cases increasingly demanded by the simulation customer base. So while the same budget buys more hardware tomorrow, tomorrow’s job will demand that additional hardware.”

“In fact, as simulation keeps getting democratized, high-end simulation will keep going ‘downmarket’ and require faster processing and quicker time to result. Whereas simulation used to be the purview of supersonic jets and Formula One, now it’s being used to design smart phones and e-book readers. This trend is, in fact, accelerating.”

Future Growth and Adoption

So we have seen that the cloud simulation market is ripe and that the technology is mature enough to deliver on certain expectations. And if the current trends are to continue, we will see simulation becoming democratized and more accessible to nonspecialists.

What does the near future hold for the industry experts that we spoke to?

“Rescale is uniquely positioned, as we support multiple applications and multiple cloud providers and hybrid on-premise data centers in a single turnkey, open-platform solution,” said Anselmi. “We believe an open platform is a better path for the ecosystem of users and ISVs relative to single-point solutions.”

“Although cloud is here to stay for HPC, this is only the beginning of the sea change,” Anselmi continued. “The advantage of the cloud is that it dynamically matches the needs of the user. As users have varying needs over time (peaks and valleys) and the need for different architectures, the cloud addresses these needs far better than an on-premise system. Over time, pay-as-you-go for hardware and software will enable users to match their needs, with no wait, improving simulation time, innovation and time to market.”

So it seems that Rescale understands the future needs of customers and software vendors who are using the Rescale platform.

What do the software vendors see in the not-too-distant future?

“First, we will continue to work with the network of cloud-hosting partners to ensure our software runs robustly and is optimized on their more than 100 datacenters worldwide,” said McDevitt.

“We will continue to invest and improve the scalability of our solvers and other computationally heavy services to make sure they take advantage of the HPC resources available in the cloud. We are also making investments in our existing products to make it far easier for our customers to leverage cloud HPC resources, and we are developing cloud-native services that improve collaboration for complex simulation workflows and the efficiency of large-scale design point evaluations. You will be hearing more about these in the near future.”

Likewise, Autodesk is also extremely proactive in its cloud strategy.

“You’ll see a lot more simulation capability appearing in Fusion 360 in the future,” said Brian Frank, senior product manager at Autodesk.

“You’re now already seeing generative design appearing in Fusion360. We will continue on that path. Simultaneously, we’ll continue to augment our simulation capabilities in other products. But at the moment, most of our simulation products are already cloud enabled. So Autodesk CFD, for example, runs in the cloud, and we just accomplished a major update of that so we can more easily take advantage of parallel computing in the cloud. Moldflow has been in the cloud for many years already. We’ve now added DoE (Design of Experiments) capability to the cloud. In general, we’ve gotten our feet wet in how to use the cloud, and now we are running as fast as we can to deploy it everywhere.”

“One of the core premises of the way that we’ve deployed simulation at Autodesk is that we believe simulation has to be part of the entire design and manufacturing workflow" added Fallon. "One of the reasons I think simulation hasn’t had the impact that it could have had on industry is because simulation is generally done by specialists outside of the product development process. The simulation tools generally don’t sit within product lifecycle management tools. You have some simulation in CAD tools, but the majority of simulation is still done externally. So we are trying to bring it into the design process.In many ways, generative design is about bringing simulation into the design process. The beauty of that is that you get a prevalidated design. So not only is the tool coming up with all possible parameters of design options for the end user, but they are valid design options."

So Autodesk are focused on bringing simulation downstream. But there is more to cloud simulation than just pure computing grunt. The cloud offers enhanced communications too.

“I think we will continue to see more of those cloud-centric features such as the collaborative workflows." continued Frank. "For example, we’ll be looking at keeping the data on the cloud instead of downloading the data to do something with it. We’d like to see that data, or portions of that data, taken from the simulations and pushed downstream [on the cloud] into operations, or generating geometry from it or information that can support manufacturing. That’s where I think we are concentrating. We’d like to say that we have simulation running well in the cloud, but now as we generate this extremely valuable information from those algorithms, how do we take it downstream and make it ultimately valuable to the customer?”

Whatever happens in terms of available platforms, Rescale, owning the largest public cloud network, seems to be perfectly positioned to supply cloud services to those who increasingly demand them.

“Overall, the most innovative and leading companies are the most advanced and heaviest users of simulation software and HPC hardware,” said Anselmi. “The digital asset that companies build with their simulation capability is always a fundamental piece of the company IP. Digital transformation and fundamental software capability are critical to success in most verticals served by simulation and HPC. We expect the leaders in these industries to continue consuming more simulation software and HPC hardware at an exponential rate.”

Instant-On

According to Autodesk, there are three main value propositions to be considered for those wishing to develop a cloud simulation strategy in their company.

“First, there’s the availability of compute, and there’s a lot of it if you need it—hence the idea of infinite compute. Then there’s the collaboration availability. So once the data is there, you can invite other people to see the data with you. The third one is the idea of instant-on technology—the idea that you don’t have to download a large file, install it on your machine and get it working. Instead, when things are on the cloud, it’s very easy to instant-on and take a look at that data. That’s where things like SimScale or SuperCool are with their instant-on capability. And that’s where Fusion360 comes in. It’s not instant-on, but it’s pretty close.”

Conclusions

So people are getting switched on to cloud simulation. But as they demand more from their simulation packages, they generate masses of data that would be more optimally used if the data remained on the cloud, rather than waiting to download huge datasets for analysis. But conservative industries are still a little concerned about data security, although cloud service providers are convinced that data is indeed safer on the cloud. These industries are going to need some convincing from service and software providers before they go all-in on a cloud strategy.It seems likely that they will jump on the cloud simulation train when the technology has matured and when their fears over security are alleviated—or when they become convinced that simulation data can be best processed on the cloud itself, removing the time to download large datasets.

Whichever comes first.

If you want to access the cloud as an individual or as a company, you can see that, via a platform such as ScaleX from Rescale, you have a variety of pay-as-you-go options for compute time as well as for use of software. Rescale does have numerous partnerships with most of the main simulation packages, so it’s not like new users will have to learn different software.

To summarize: Cloud-based simulation is cheap, accessible and familiar. It’s all of your favorite software, supercharged and on demand.

What’s not to like about it? Expect to see an uptake in cloud-based simulation customers as the technology matures and moves in from the fringes.

Next stop: productivity!