Automotive manufacturers are exploiting cloud resources for computationally demanding development tasks, as Dean Phillips, worldwide technical leader for automotive at Amazon Web Services, explains how to process simulation workloads of over one million cores
Cloud-based resources are revolutionizing the way engineers access computing power. Rather than having to house HPC (high-performance computing) hardware in-house, with the associated support staff and infrastructure demands this entails, it is now possible to access an à la carte menu of capabilities from a whole host of suppliers.
As Dean Phillips, worldwide technical leader for automotive at AWS, explains, “We offer automotive solutions across the full ecosystem – from car production to selling cars to cars on the road. We’re the only cloud provider that can offer fully integrated, end-to-end solutions for its customers, from the initial ideation and innovation on a new vehicle through engineering and manufacturing to consumer sales and after-sales.”
While automotive manufacturers invariably have their own in-house HPC resources, these will always have a finite capacity; with the ever-growing demands from areas such as EV and AD/ADAS development, being able to rapidly access on-demand computing is an attractive option for companies.
“AWS offers three strategic workloads for the automotive industry,” Phillips continues. “The first
is AWS Connected Vehicle Solution, which enables automotive manufacturers to build server-free IoT applications that gather, process, analyze and act on connected vehicle data, without having to manage any infrastructure.
“It also provides a full suite of services to support ADAS and autonomous vehicle development and deployment. OEMs are leveraging AWS to process simulation workloads of over one million cores on Amazon Elastic Compute Cloud as well as dev-ops for automotive software development and software-defined vehicle initiatives.
“Lastly, AWS supports digital customer engagement workloads with customer solutions like ZeroLight that leverage a broad set of virtual GPU instances on the Amazon Elastic Compute Cloud.”
Specifically in relation to simulation work, one of AWS’s big selling points is that it can offer in-house technology to increase efficiency and shorten development times in a cost-effective manner. It not only provides HPC for areas such as aerodynamic simulation, but also machine learning capabilities for CAD and structural simulation. In applications such as these, it is rare for users to interact directly with AWS; instead, the big players in computer-aided design/manufacturing and simulation, such as Autodesk, provide access to Amazon’s cloud computing resources through their own in-built interfaces.
However, Phillips notes that capabilities can also be directly integrated with manufacturers’ existing IT infrastructure: “This includes the ability for automotive companies to transfer massive amounts of data from real-world tests and use AWS cloud computing power to run computer simulation and deep-learning exercises at an accelerated rate. As auto makers move from simulation to assembly, they can then apply AWS analytics and Amazon SageMaker to inform testing and influence vehicle design for performance and efficiency.”
He flags up the example of Nissan using Rescale (a specialist that has developed an application-agnostic interface giving access to cloud infrastructure) and AWS HPC to reduce time-to-market for its computer-aided engineering workloads such as crash simulation and aerodynamics development.
Another area in which AWS is being deployed by many manufacturers is cataloging and interrogating the vast amounts of data their organizations generate. AWS IoT is able to provide secure connectivity and management of millions of vehicles and devices globally, in addition to tools to easily track and govern data access rights, and security updates.
For example, BMW Group used AWS to help its digital transformation with the manufacturer’s Cloud Data Hub, which processes and combines anonymized data from vehicle sensors and other sources across departments, making it easily accessible for internal teams creating customer-facing and internal applications. “The centralized AWS-based data lake forms the BMW Group’s foundation to develop data-driven IT solutions and enables the company to automatically and independently scale on a serverless architecture,” Phillips outlines.
A data lake is where largely unstructured data is stored in a flat architecture with each item of data assigned an identifier and metatags, a useful means of arranging data on everything from consumer habits to manufacturing and testing activities. In the case of a company using AWS for its data lake, much of the heavy lifting of establishing and managing such a resource is removed. “It can therefore innovate faster than it could with the previous on-premises solution, which required infrastructure management and capacity planning for each new initiative.”
Another example that Phillips details is that of autonomous driving startup Torc: “Ahead of the launch of its fleet of next-generation self-driving test trucks, Torc selected AWS to handle the scale and speed needed for data transfer, storage and compute capacity. Because of AWS’s ability to provide rapid, secure data transfer, intelligent tiered storage, managed orchestration and analytics tools and high-performance multi-core CPU and GPU compute, Torc was able to scale its agile and cost-efficient development platform and accelerate its testing and commercialization of the technology.”
Of course, AWS is not the only player on the cloud compute block, and even with a smaller market share, Microsoft, with its Azure platform, is also heavily invested in the automotive market. Regardless of supplier, the advent of easily scaled, rapidly accessible cloud computing resources can only be a boon for automotive manufacturers in the increasingly data-heavy and simulation-led automotive industry.
Rivian: AWS publicized in 2021 that electric vehicle manufacturer Rivian harnesses the cloud to perform powerful simulations, with AWS enabling speed and scalability – reducing the need for real prototypes. According to AWS, the OEM was previously limited by the compute capacity of its on-premises compute infrastructure. Rivian migrated to a new stack and used various AWS services. As AWS notes, using Amazon EC2 C5n Instances, Rivian’s software performance has improved by up to 66%. Engineers can reduce their focus on managing the technology and instead home in on development tasks.
Keyou: In 2021 AWS highlighted how technology startup Keyou turned to AWS to develop a new computational fluid dynamics simulation tool. The solution is said to be both fast and scalable, capable of supporting hundreds of cores and able to produce results in just a day.
Volkswagen: In another of its 2021 case studies, AWS detailed how Nice DCV, a high-performance remote display protocol included within AWS, has enabled Volkswagen Passenger Cars to stream remote applications to 1,000 engineers, providing near-real-time responsiveness and enhanced security.
BMW Group: In late 2020, it was announced that AWS and BMW are jointly developing cloud-based IT solutions and upskilling up to 5,000 software engineers in cloud technology.
This article was first published in the November 2021 issue of Automotive Testing Technology International