Nissan Motor’s engineers rely on large-scale computer-aided engineering (CAE) and high performance computing (HPC) during the car design and development stages. But in keeping with its goal of achieving carbon neutrality and zero-emission vehicles by 2050, the company decided that it needed a more energy-efficient way to generate that massive computing power.
Historically, Nissan, which celebrated its 90th anniversary on December 26, maintained its own on-premises data centers to power its CAE and HPC activities. But the energy consumed by that infrastructure became a potential barrier to its commitment to reducing CO2 emissions.
“Energy consumption in our data centers was a major concern for us,” says Dai Matsubara, a senior manager in the company’s engineering and design systems department. “Plus, large fluctuations in demand for computing resources, typically occurring during the transition of product development projects, meant that these expensive computing resources sometimes operated at low capacity.”
Consequently, Nissan’s leadership evaluated ways to migrate development and testing activities to more flexible and scalable cloud computing environments that would enable technical staff to respond quickly to changing demands while also reducing energy consumption.
Initially, Nissan’s technical team tested design and development activities in a hyperscale cloud computing platform offered by a provider using a virtual machine (VM) strategy. However, the virtualization layer created performance problems for the large-scale simulations, which sometimes exceed 1,000 parallels simultaneously.
Consequently, Nissan’s technical staff turned to Oracle for a proof of concept (POC) of bare metal instances on Oracle Cloud Infrastructure (OCI). Unlike with virtualized instances, OCI Compute bare metal instances gave Nissan all the compute power of a fully dedicated server with no virtualization overhead, delivered on demand for various customized applications.
“We chose HPC on OCI with bare metal high performance instances, as part of our cloud-first/multicloud strategy, to meet our global challenges of daily increasing simulation demand while meeting constant cost reduction demands. As expected, Oracle Cloud HPC brought Nissan a great return on investment.”
OCI is one of the first public clouds to offer bare metal HPC in a remote direct memory access (RDMA) cluster network. The POC validated that OCI could surpass Nissan’s performance expectations, with pricing comparable to the VM services it previously tested with the other cloud provider.
Nissan’s technical staff worked with Oracle throughout the migration, from defining requirements and customizing functions to full project implementation. “The migration process was extremely smooth,” Matsubara says. “Our internal teams quickly established a great working chemistry with Oracle’s support staff. The project took only about two months, and we didn’t experience any major problems.”
Nissan has migrated more than 40 customized CAE applications used by its engineers globally to OCI Compute bare metal instances.
Workloads at peak times can require tens of thousands of cores, which deliver superior performance and flexibility at a lower cost than its on-premises data centers. NVIDIA A10 GPUs, employed by the virtual desktop infrastructure for each CAE application, deliver 100 Gb/sec bandwidth which results in minimal latency.
OCI's autoscaling adjusts the number of Nissan’s compute instances to help deliver consistent performance for end users during periods of high demand, while also reducing costs during periods of low demand. Technical staff can automatically start up the necessary nodes and shut them down as soon as they’re no longer needed.
The significant performance improvement of highly parallel computing, powered by OCI, not only speeds up various simulations and improves the productivity of engineers but it also contributes to reducing application license costs. Some of the CAE applications Nissan uses are billed at an hourly rate, and those license fees vary depending on the number of allocated CPU cores. With OCI Compute bare metal, the number of simulation runs has increased but the license costs have remained the same. Overall, the OCI solution is about 30% less expensive than Nissan’s previous on-premises environment.
“We are already exploring ways to use AI and ML to shape the future of sustainable mobility and innovative designs,” Matsubara says. “We will rely heavily on Oracle and OCI to help us arrive at our desired destination.”
Auto manufacturers use cloud technology to reduce the frequency of physical testing.
Learn why HPC on OCI is the right solution for your HPC workloads.