DEHN: Lightning strikes the cloud in Microsoft Azure

The challenge: A local simulation center is too expensive

As one of Germany’s “hidden champions,” DEHN knows the value of its in-house expertise. This is why DEHN, an SME based in the Bavarian city of Nuremberg, carries out almost all of its product development activities at its location near Neumarkt. To stay ahead of the international competition, the company decided to simulate not only phenomena such as voltage fluctuations, but also the tremendous power of full-blown lightning strikes. In order to put this idea into practice, DEHN enlisted the help of Dr. Martin Hannig, an expert from the Technical University of Darmstadt, as development engineer. He was originally brought in to oversee the creation of an on-site simulation center. 

However, it quickly became clear, that this endeavor would be both complicated and expensive. Dr. Hannig’s research team determined that the sheer complexity of the simulations called for a high-performance computing (HPC) cluster as well as additional HPC resources to visualize the results. But DEHN didn’t have the know-how required to procure and operate such equipment and was unsure whether it could justify the high cost of investment. After all, this kind of local simulation center would tie up a great deal of money over several years. And DEHN would have to scour the all but empty labor market to find suitable experts to run it.

The solution: High-performance computing in the Azure cloud

As a long-standing Microsoft customer, DEHN was already putting time into determining the potential of Azure for various cloud workloads as part of its hybrid cloud strategy. It was at this point, that the company turned to Microsoft to find out if complex mathematical and physical problems, such as those associated with simulating lightning strikes, could also be calculated in the cloud. For this ambitious undertaking, Microsoft brought in one of its Global Black Belts as well as UberCloud, a consulting company specializing in cloud-based HPC and digital twins. Within a few short weeks between November 2020 and January 2021, the project team had worked up a proof of concept. As a result, carrying out the complex and expensive simulations in Azure is not only possible, but for SMEs like DEHN it is also more practicable. 

“HPC simulations for lightning strikes might seem like a complicated way to start our migration to Azure,” says Manfred Ferstl, Head of IT Infrastructure at DEHN. “But compared to an on-premises solution, the complexity—and with it the effort—was much lower for us. This is because the discussions with Dr. Hannig’s team and implementation of the technical details were handled mainly by UberCloud.” 

Together, the project team of DEHN, Microsoft, and UberCloud created an HPC environment based on Azure HPC, Azure CycleCloud, and other modules. This environment allows DEHN’s team of engineers to carry on using their tried and true simulation solutions, which saves time and training costs while offering a previously unknown level of flexibility. When it comes to establishing and scaling the environment currently needed, experts from the research division take the lead: “Azure’s self-service principle saves us from having to intervene in simulation runs,” Ferstl reports. “Dr. Hannig’s team can simply create the desired HPC environments itself—naturally always according to our guidelines and within the approved budget.” This method allows a simulation environment to be set up within minutes, and then dismantled as soon as it is no longer required. “Thanks to the HPC environment in Azure CycleCloud, we’re in a position to draft cluster configurations that perfectly match our needs in next to no time,” Dr. Hannig says. 

“Using Azure HPC, we can connect up to 1,000 computational nodes if we need to, and in turn greatly accelerate simulations,” Ferstl explains. “The only question is what level of acceleration is worth what level of cost?” Here, the project team has succeeded in striking a balance in which simulations take between a few hours and two days. “Azure HPC allowed us to accelerate our simulation by a factor of six compared to running them locally—without having to invest in additional hardware and at very bearable operating costs,” Ferstl says. A high level of data security is ensured throughout, a must for this company: “Even with the Azure-based solution, our data doesn’t pass through the public internet,” Ferstl says. “Instead, it’s integrated into the company’s own private network.”

The research team finds working with the solution easy. “The intuitive user interface was a big help, for example when setting up additional nodes for an expanded parameter study in which we’re examining electric arc plasmas,” Dr. Hannig reports. In addition, the team of engineers no longer needs expensive workstations for visualization: the graphics power is supplied by virtual computers in the cloud.

Ferstl was thrilled with how the project went: “Microsoft wasted no time in bringing in UberCloud, which gave us exactly the HPC experts we needed,” Ferstl says. “This minimized our migration and implementation costs.” For the next step, he wants to move evenmore workloads to Azure to make the self-service principle and lightning-fast scaling available to other in-house teams as well.

Thanks to the cloud-based simulation environment, DEHN is saving time and the money it would otherwise spend on highly paid experts in research and IT operations. The research division can run simulations faster and achieve more precise results. This in turn allows the manufacturer to shorten its product design phase and bring innovative products to market faster. In other words, there’s nothing standing in the way of DEHN continuing its legacy as an innovation leader. In hindsight, the decision to switch to the Azure HPC environment has proved to be a real flash of inspiration.