AWS’ SDV Chief Explains How the Cloud Is Revolutionizing Your Car
Get In: The Software-Defined Vehicle Podcast from BlackBerry
![](/content/dam/blogs-blackberry-com/images/blogs/2024/01/get-in-aws-875x530px-s2-e8.png)
In sub-atomic space, particles get excited by interacting with each other at close range. The energy contained in one particle essentially rubs off on its close companions, and that excitement becomes contagious. The same thing happens when you start bouncing ideas around with people like Stefano Marzani, worldwide tech leader of software-defined vehicles at Amazon Web Services.
“Cars are changing radically. There's a disruption right now in the sector,” Marzani told me during a recent, highly charged interview at the LA Auto Show. He proceeded to explain exactly what’s causing this radical disruption, and some of the most important trends and takeaways for 2024 and beyond.
Welcome to Season 2, Episode 8 of “Get In: The Software-Defined Vehicle Podcast from BlackBerry.” This series explores the possibilities created by — and technologies behind — the revolution in global transportation we are witnessing today. In this episode, I speak with Marzani about how the cloud is poised to “hyperscale” our cars, imbuing them with new intelligence and capabilities that are starting to sound like superpowers.
Click below to watch the full Season 2, Episode 8 podcast.
Going for the Gusto
Like the global powerhouse that employs him, you quickly get the sense from Marzani that he isn’t “messing around”: He seeks out the most intractable problems and goes straight at them with world-class technology and resources, to find solutions. As the bio for his own podcast series at AWS asserts, he is “focused on solving the biggest challenges in the auto sector” — from building secure autonomous vehicles to creating new mobility fleet solutions, by applying deep expertise in IoT, machine learning, vehicle architecture, and automotive software development & tooling.
Spending even a few minutes with Marzani gives you the feeling that those big challenges haven’t got a fighting chance.
Functional Safety, Security, and Reliability at Scale
“The cloud is technically not part of the car itself,” explains Marzani. However, the dividing line between where the car ends and the cloud begins is getting decidedly hazy.
“All the OEMs that we're working with are thinking about how they can overhaul the electronic architecture, consolidating the more than 100 ECUs (electronic control units) that they have in the car right now to a few very powerful ones. But you can't just take the software that is inside those boxes and plop it into a consolidated unit. It will simply not work.”
This is clearly one of those big challenges that Marzani and his team are in the process of solving, with help from certain key partners.
“You must reorganize the software stack around 10 to 15 or more cores inside the automotive computer. You must partition that into different operating systems in different processes.”
It’s a difficult proposition, “But this is exactly what we do in the cloud,” Marzani explains. “If you take one of our AWS Graviton processors that are ARM-based and have 64 cores, you partition the 64 cores.” For example, “Give four of them to an OS for a specific purpose, and another three for another OS for another specific purpose. These could be microservices and cloud-native technologies.
“This is the technological idea that is very useful for reorganizing the automotive software stack.” But that’s just for starters.
“Then there's the cloud, the ‘big data center’ environment. Already the automotive sector uses that extensively for storing the data that is coming from vehicles in the test fleet or connected cars. It’s also used for simulations, verification, and validation activities for new software design testing, and machine-learning training, especially for ADAS or autonomous driving functionality.”
Automotive Virtualization and AI to Speed Development, Reduce Cost
What’s next? The immediate future has a lot to do with virtualization and artificial intelligence, things that require levels of compute power that may not yet be feasible or cost-effective to house in a moving vehicle, but which are readily and abundantly available in the cloud. “AWS is a framework that enables you to build some very relevant features,” says Marzani. “One is virtual sensors that can be deployed inside vehicles to get insights.”
Another example is what these virtual sensors can enable when AI is applied to the resulting data. “For example, you could deploy an intelligent algorithm that can monitor the behavior of the high voltage battery across the EV fleet to provide recommendations on how to optimize range.” Machine-learning algorithms working with sensors can provide these kinds of insights, particularly for enhancing safety. “That's the beauty of the collaboration with BlackBerry, which is very well known for security, functional safety, and reliability. You can really go deeper inside the vehicle and distribute this content even on critical systems.”
Virtualized sensors are only the tip of the iceberg, however. Why not go ahead and virtualize the rest of the vehicle, starting with key hardware components and subsystems that developers need for testing and fine-tuning their software — but which are typically in extremely short supply during the critical design stage of manufacturing?
“The QNX Hypervisor is very popular on the market as a virtualization mechanism,” Marzani confirms. One major reason: It solves the problem of limited availability of physical test hardware during development. “That is a pivotal moment for feature development because you're going from a scenario where you need a piece of hardware on your desk to work, to a scenario where if you have a browser, you can work as a developer, because the systems will be available through the network in a virtualized form in your browser,” Marzani enthuses.
“For example, this is being used for our product development with an Android-based infotainment system,” continues Marzani. “This operating system is fully virtualized. We have presented this model in production with a couple of customers already and the effect of it is total disruption. The first version of the instrument cluster infotainment system was shown at CES 2023.
“Adopting this virtualization technique means everybody in the world can now develop for that specific component — even if they don't have the component. This can deliver a 70% reduction in time and a 50% reduction in costs, especially in the prototyping phase. You really can create a community of developers, but you can’t if you still rely on a physical component.”
Vive la Révolution!
Could this be the beginning of a revolution in automotive design, akin to the arrival of smartphones and ride-sharing apps like Uber?
“I really hope so,” says Marzani. “The reason why it hasn’t happened yet in cars is that they are a delicate product. If you do something wrong, you really risk the life of the people in that car.” This is a core reason AWS teamed with BlackBerry and its QNX® real-time OS, culminating in the joint development of the BlackBerry IVY® platform for cloud-enabled, software-defined vehicle designs.
“Automotive is a sector that is perfect for BlackBerry because of its certified operating system hypervisor. On the powertrain, on brake-by-wire systems, and on steering, you don't want to have problems with software. You must take that into consideration when you create features for the car such as ADAS,” or advanced driver-assistance systems, Marzani explains.
“We’ve got to do it right, to provide the developers not just the tools and APIs, but also ways to test the software in real conditions.”