Manufacturers have been using simulation to envision the look, feel, and performance of their product for more 30 years. Digital product design pervades the manufacturing industry, from high tech to automotive, from consumer goods to aerospace. The digital design process involves creating a digital model and then testing it for failure. Engineers aim to replicate physics, then use that data to optimize the design.
Figure 1: Simulation-driven process for digital design. Image provided by Altair.
With machine learning, simulation-driven design has entered a new chapter: predicting event outcomes based on historic design data. These techniques help simulate autonomous vehicles using advanced driver-assistance systems (ADAS).
Around the world, manufacturing companies face challenges from regulatory agencies, for safety and environmental concerns, and evolving consumers. In the automotive industry, electrification and driverless cars are a reality and continue to accelerate.
Addressing these challenges means significant investment in skill and technology. For example, automakers use simulation-driven digital product design to accelerate time-to-market. That process includes using crash simulation techniques to test for the following qualities:
Noise vibration harness (NVH)
More robust designs lead to fewer recalls and failures in the field. Designing products while simulating physical scenarios requires a tremendous amount of computational capacity. It takes hundreds of thousands of compute cores to solve millions of simulation problems. The faster the simulation runs, the faster an automaker can change in response to new demands or requirements. Faster iterative improvements also lead to better designs. It doesn’t do the company any good if compute resources aren’t available on-demand and highly paid engineers are waiting for their turn.
A large ecosystem of independent software vendors (ISVs) offer simulation applications that run on HPC infrastructure, and their licensing fees can range in millions of dollars. So, it’s imperative that organizations run on state-of-the-art HPC infrastructure, so they can get their money’s worth.
Many automakers run thousands of HPC cores in on-premises data centers. They update this hardware every three years in an arduous process that can take up to 12 months. While the rest of the world moves on to new advancements, this model also locks automakers into a technology stack for CPU, networking, and storage. Computing demand for digital product design HPC by automakers continues to grow more than 20% per year, but on-premises data centers aren’t equipped to handle this increasing demand without a huge upfront capital investment. In contrast, cloud computing with GPU acceleration provides the flexibility, performance, and capacity that automakers want.
Figure 2: Cost breakout for cloud compared to on-premises computing.
In the previous chart, we suggest a new paradigm for structuring HPC compute capacity in the cloud. Instead of capital investment for on-premises fixed capacity, customers are envisioning a different cost structure: fixed capacity in the cloud with flexible capacity as needed. Oracle’s annual flexible commitments and Universal Cloud Credits make it easier for customers to plan and expand their HPC infrastructure on-demand as an operational expense.
Traditionally, automakers have built and maintained their own compute capacity for simulation tasks. With Oracle Cloud Infrastructure's innovative bare metal HPC compute offering, cost-effective simulation is finally here. Further, a large Japanese manufacturer moving to Oracle Cloud Infrastructure for digital product design is having a ripple effect throughout both the discrete manufacturing industry and the ISV ecosystem. Our customer wanted to keep up with increased simulation demands, while keeping their costs in check without expanding on premises capacity.
This customer saw numerous technical and business benefits in selecting Oracle Cloud Infrastructure:
Oracle Cloud Infrastructure’s industry-first Intel Xeon based bare-metal compute infrastructure with RDMA cluster networking offers latencies of under two microseconds.
100-Gbps RDMA bandwidth enables HPC simulations at scale on Oracle Cloud Infrastructure.
Instances with NVIDIA V100 Tensor Core GPUs for faster, more cost-effective structural simulation, compared to on-premises or other cloud providers
Reduced the overhead of large data transfer, while maintaining that all simulation data viewable in 3D OpenGL format
They anticipate higher performance and lower costs with the ability to easily run their engineering simulation workloads in the cloud.
Try Oracle Cloud Infrastructure and see how it can help your organization design robust products.
Every use case is different. The only way to know if Oracle Cloud Infrastructure is right for you is to try it. You can select either the Oracle Cloud Free Tier or a 30-day free trial, which includes US$300 in credit to get you started with a range of services, including compute, storage, and networking.