Advanced Computing in the Age of AI | Friday, March 29, 2024

Altair Pushes CAE Boundaries with Private Cloud Appliance 

<img style="float: left;" src="http://media2.hpcwire.com/dmr/hyperworksunlimited2.jpg" alt="" width="95" height="69" border="0" />In an effort to knock down the barriers to the adoption of simulation-based design via high performance computing (HPC), Altair has launched HyperWorks Unlimited, a private cloud appliance that bundles HPC hardware with the company's solvers and visualization tools.

Each year, more and more manufacturers discover and embrace the benefits that simulation-based design can offer their products over traditional computer-aided design. Unfortunately, employing advanced simulations is not nearly as simple as paying for the software. An investment of this kind comes with it a need for high performance computing (HPC) infrastructure, as well as experts for both the software and hardware

But with the launch of HyperWorks Unlimited, Altair hopes to knock down two of these obstacles for manufacturers looking to jump on the HPC-powered train. Going beyond their traditional software offerings, Altair has stepped up to provide a fully configured appliance designed specifically for computer-aided engineering (CAE).

Once users bring the managed private cloud into their offices, they gain unlimited used of all Altair software including HyperWorks applications and PBS Works workload management tools.

“We have reached an inflection point in product design where hundreds of thousands of mid-sized manufacturers are now looking to be even more competitive by moving beyond traditional computer-aided design environments and into simulation-based design,” said Stephen Wheat, general manager of the High Performance Computing Group at Intel. “This requires a tremendous amount of compute power that’s easy to deploy, manage and use.”

But as many manufacturers are already aware, making the jump into HPC hardly that simple. Lack of expertise for using and maintaining HPC configurations, cost barriers to HPC, the high cost and low scalability and CAE software each represent costs that can be hard to justify.

In a recent sit-down with Digital Manufacturing Report, Sam Mahalingam, Altair senior vice president of enterprise products explained that these were the issues, or “pain points” that HyperWorks Unlimited was designed to tackle. “We basically went back to the drawing board and asked, ‘How do we address these two pain points for small to medium businesses,” he said. “With this intent in mind we said, ‘Okay, let’s come out with a bundle that is a hardware-software solution that would alleviate some of these customer pain points.’”

For companies that looked to bypass these HPC cost barriers through outsourcing, Altair had already tackled cloud-based simulation with HyperWorks On-Demand, but as Mahalingam pointed out, license scalability remained an issue, but data security and the protection of intellectual property was the chief concern when dealing with public cloud services.

“In order to alleviate this, we said ‘What if we came up with a private cloud solution that’s completely on their private network so that the data, the licenses for the software, as well as the hardware is all on their premises?’” he explained.

To make this a reality, Altair looked at their own experience with HPC to determine what the optimal hardware vendor and configuration would be for SMMs. Although Mahalingam said that all top hardware vendors were in consideration from the start, SGI’s extensive experience and history in the HPC and CAE spaces along with its hardware’s ability to be densely packed to place the greatest number of compute nodes in a single appliance brought it to the top of Altair’s shortlist.

After the hardware was chosen, Altair’s next challenge was to integrate all of the solvers and visualization tools this in a way that wouldn’t limit the end-user in terms of the number of jobs they could run on the hardware provided.

“Based on the hardware that is given to them, they can run a single job and utilize all the cores, on they can run many jobs using a couple of cores each,” Mahalingam said. Meanwhile, the upper limit of varies based on the hardware configuration that a customer purchases from Altair. On the smaller end you have four compute nodes and four graphics nodes that total up to 96 cores. Mahalingam noted that for small companies, the median job utilizes 16 cores, which means that even on the most basic offering a company could expect to run up to 6 jobs at a time.

And while these more humble configurations may seem meager next to traditional HPC systems, Mahalingam was quick to point out just how much SMMs could save when opting for Unlimited.

“If a small company goes and invests in their own HPC infrastructure, I would say that it would be very expensive in terms of total cost of ownership, because at the end of the day it’s not just the hardware they’re investing in but the personnel to maintain this hardware  and additional software costs they’re going to incur,” Mahalingam said. “If you look at it from a percentage perspective, I would say it would probably be three to four times more expensive for their own [HPC] investment rather than procuring Unlimited and working with it.”

EnterpriseAI