Oracle on Tuesday announced that Nissan is migrating on-premise,high-performance computing (HPC) workloads to Oracle Cloud, in order to perform latency-sensitive engineering simulations.
Back in 2018, Oracle introduced bare metal compute instances, powered by Intel Xeon processors, tailored for HPC workloads. The instances are part of Oracle's "Clustered Network" offering, which provides access to a low-latency, high-bandwidth remote direct memory access (RDMA) network. Nissan is one of the first automotive OEMs to leverage Oracle's bare-metal
GPU-accelerated hardware for HPC workloads.
Bing Xu, the GM of Nissan's Engineering Systems Department, said the company selected Oracle's cloud HPC offerings "to meet the challenges of increased simulation demand under constant cost savings pressure."
Specifically, Nissan uses software-based Computational Fluid Dynamics (CFD) and structural
simulation techniques to design cars. These techniques help the company test a car design's external aerodynamics and potential structural flaws -- factors that impact their fuel efficiency and safety. Large CFD simulations require tremendous amounts of compute power.
Tapping into the Oracle cloud, Nissan will have the flexibility to launch tens of thousands of
cores and GPU-based visualization servers when needed. The data generated by simulation workloads will be easily viewed in 3D OpenGL format in the cloud.
As of late 2019, about 20 percent of HPC workloads were running in cloud environments, according to the research firm Hyperion. About two years prior, only 10 percent of HPC workloads were running in the cloud. Moving HPC workloads to the cloud still faces hurdles, Hyperion noted, such as the high costs associated with data locality when large volumes of data are involved.