Download PDF

Altair > Case Studies > Subaru's Migration to Cloud-Based High-Performance Computing for Enhanced Vehicle Safety

Subaru's Migration to Cloud-Based High-Performance Computing for Enhanced Vehicle Safety

Technology Category

Infrastructure as a Service (IaaS) - Hybrid Cloud
Infrastructure as a Service (IaaS) - Private Cloud

Applicable Industries

Automotive
Cement

Applicable Functions

Quality Assurance

Use Cases

Manufacturing Process Simulation
Transportation Simulation

Services

Cloud Planning, Design & Implementation Services

The Challenge

Subaru, a global automobile and aircraft manufacturer, is committed to achieving zero fatal traffic accidents by 2030. To reach this goal, the company needs to continually innovate and ensure high collision safety, which requires conducting Computer-Aided Engineering (CAE) simulations using High-Performance Computing (HPC). Subaru had been maintaining its own HPC environment near its main manufacturing facility in Japan’s Gunma prefecture. However, as the computational processing requirements for simulation increased, the team faced a shortage of power and space for expansion. They started using a private cloud located in a remote data center in Tokyo, which required a dedicated line for user access. Due to the high cost, they decided to evaluate public cloud options and sought recommendations from the Japan Automobile Manufacturers Association’s cloud working group.

About The Customer

Subaru is a global manufacturer of automobiles and aircraft, with a strong focus on safety and human-oriented manufacturing. The company's roots are in aviation, but its automotive division has evolved to become its core business. Subaru is committed to delivering value to its customers by continually improving and refining its designs. The company's goal is to achieve zero fatal traffic accidents by 2030. Over the past decade, Subaru has already halved the number of fatal and serious injuries per million vehicles sold. To further this initiative, Subaru must continue to innovate and ensure high collision safety, which requires conducting CAE simulations using high-performance computing (HPC).

The Solution

After evaluating various factors such as cost, performance, security, ease of use, and expert support, Subaru chose Oracle Cloud Infrastructure (OCI) with workload management by Altair® PBS Professional®, a solution they were already familiar with and trusted. PBS Professional comes with a built-in GUI that allows users to extend their HPC resources and utilize dynamic cloud bursting for peak-time workloads. Subaru confirmed that calculation processing wouldn’t be compromised by a move to the cloud, and they optimized their new environment to ensure efficient performance. The migration of its collision and fluid simulation environment to OCI was a large-scale operation completed in around six weeks, significantly shorter than the 3-4 months it would typically take to expand an on-premises HPC environment. The team also reduced the amount of data transfer required by setting up remote preprocessing.

Operational Impact

The transition to Oracle Cloud Infrastructure (OCI) has been beneficial for Subaru. Their CAE users have been satisfied with the change, and performance optimizations like increasing per-node memory capacity have been implemented. The team also reduced the amount of data transfer required by setting up remote preprocessing. Throughout the transition process, Subaru stayed tuned in to user needs by listening to feedback and thoroughly testing new procedures. Now, Subaru is expanding its cloud HPC resources and using hybrid cloud for efficient, flexible, and cost-effective CAE to enable continuous vehicle safety innovation. The company appreciated the expert technical support from Altair and Oracle during the transition.

Quantitative Benefit

Subaru reduced the cost of HPC resources by hundreds of millions of yen.
The effect of reducing labor requirements for private cloud operation and management was significant.
The migration to OCI was completed in around six weeks, significantly shorter than the 3-4 months it would typically take to expand an on-premises HPC environment.