How Virtualization Solves Five Automotive OEM Business Challenges
Virtualization — a key approach in enterprise IT architecture — is expanding into the world of embedded systems. The automotive industry sees the benefits of virtualization with increasing clarity as vehicle onboard systems become more sophisticated and connectivity ubiquitous. Business challenges evolve, and new ones emerge, for automotive OEMs as they work to deliver the cars of tomorrow – better known as software-defined vehicles (SDVs).
The emergence of embedded hypervisors plays a crucial part in meeting five key challenges that automotive OEMs face today:
Challenge 1: Reducing Hardware Complexity and Cost
Chip shortages continue to plague the industry, and the latest forecasts indicate this issue will linger well into 2023. This is a major factor accelerating hardware consolidation onto single system-on-chip (SoC) designs.
More strategic, however, is the evolutionary path of ECU (electronic control unit) consolidation – leading to domain-driven vehicle architecture and ultimately, to the holy grail of in-vehicle high-performance computing (HPC) platforms. The road to onboard HPCs presents a complicated journey for OEMs. While it will yield a decrease in the number of hardware components, thermal dissipation, and supply-chain dependence — as well as savings in development, testing, and toolchains — it means the software needed on the SoC becomes significantly more complex.
The SoC needs to perform all the capabilities of the systems consolidated onto it, just as if they were separate systems. To do that, we need to run each component’s discrete operating system (OS) simultaneously, with each one safely isolated on its own virtual machine (VM). This is the role of the embedded hypervisor, running on top of an embedded real-time operating system (RTOS).
Further value can be realized when one considers the ability to easily port legacy code to the latest hardware, effectively re-using software across multiple vehicle platforms. This is all possible via virtualization.
Challenge 2: Increased Emphasis on Functional Safety
The automotive industry has a strong safety culture running through its engineering and manufacturing operations, so it understands that safety must be systemic. It can’t be simply “bolted on” at the end of any process.
The consolidation of multiple discrete systems on a single SoC brings with it specific safety requirements, emphasizing the importance of “safety by design.” For example, managing safety-critical instrument clusters alongside infotainment-related functions means enforcing strict isolation of one system from the other — a key principle of ISO 26262. Indeed, if advanced driver assistance systems (ADAS), anti-lock braking systems (ABS), power steering, or other critical systems are involved, then the highest degree of certification – ASIL (Automotive Safety Integrity Level) D is needed. This requires the hypervisor to guarantee isolation of safety-critical processes, ensure safe inter-process communication, and provide localized recovery of failed applications — all with high availability and reliability, and covering all systems running on the SoC.
Safety can never be compromised, so it must be commercially cost-effective to achieve. This necessitates streamlining the certification process as a key element. Certifying only those elements of the system that are safety-critical and developed in-house — excluding non-safety-critical elements running on Android™ or Linux® OSs, or toolchains used — leads to significant savings in time and cost. Therefore, using an embedded hypervisor that offers the highest safety certifications out-of-the-box is a must.
Challenge 3: Cloudification
The term “Cloudification” has achieved growing popularity in the automotive industry, as it adopts cloud-based approaches and tools that hold the potential for cost savings. To understand how a hypervisor plays in this arena, let’s break down what cloudification means to vehicle makers, or OEMs (original equipment manufacturers).
Cloudification in the Lab
Running OSes in the cloud (such as Android, Linux, or QNX®) allows developers to work on algorithms independently from their target hardware. Bringing these together, still in the cloud, is where a cloud-based hypervisor comes in. It allows each OS to run in its own respective VM, effectively simulating target hardware. The software is then taken from the cloud and launched on the actual target hardware, via a connected target running an onboard hypervisor.
The use of a hypervisor in the cloud, acting in a software-as-a-service (SaaS) capacity, brings significant value for OEM software development teams, by doing the following:
- Reducing “developer friction” — anything that causes a “drag” effect on efficiently performing daily work. Time and effort spent on procuring target hardware, or learning/setting up tools, creates friction. Reducing cognitive load on a developer allows more time to solve problems and create market-differentiating features. Cloud-based tools that are kept up to date by the vendor, need no installation, have rapid ramp-up, and offer additional developer seats, may be purchased easily via a cloud provider marketplace — an approach that’s likely familiar to most teams.
- Accelerating software development cycles. This de-risks schedules, improves time to market and reduces costs. The term “shift left" (moving development and testing to the left of the timeline, thus earlier in the process) is gaining traction in the automotive industry. Development and testing can be integrated more easily within a continuous integration/continuous delivery (CI/CD) pipeline — a familiar concept in modern software engineering circles, which OEMs are adopting rapidly.
Cloudification in the Vehicle
Massive amounts of data moving between the vehicle and cloud-based systems will only trend upward with autonomous vehicles, as machine learning (ML) algorithms running in the vehicle are trained using big data sets via associated processing power in the cloud. Additionally, new automotive business models (ridesharing, insurance) that require vehicle-to-cloud connectivity must be facilitated. The concept of “blending” between the edge (aka, the vehicle) and the cloud — in terms of data transfer and shifting computing resources — presents an increasingly important facet of cloudification.
With an SDV generating one to two terabytes (TBs) of data on an average day of driving — much of which is processed on-board, with a subset sent to the cloud as necessary — the power of an embedded hypervisor becomes significant for facilitating vehicle domain-based architecture and the HPC.
Challenge 4: Vehicle User Experience as a Market Differentiator
Vehicle buyers consider their new vehicle an extension of their digital selves. Similar to smartphones, users have become accustomed to ease of use, near-constant connectivity, and innovative features at their fingertips. And they expect the same from their vehicle, while not compromising safety. This presents OEMs with challenges, but also enormous opportunities to differentiate in a competitive market.
With the emergence of ridesharing, last-mile delivery, and other fleet-oriented business models, OEMs must also compete for fleet procurement contracts that rely on customized features and capabilities. In addition, they need to provide specific vehicle functions that become “table stakes” for staying competitive, like ADAS or acoustics system management features.
One way to achieve these ends is to leverage a robust hypervisor that can support complex deployments, such as a legacy Linux application stack running in parallel with a newer Android stack and a QNX stack on the same SoC — all without having to re-architect or rebuild the core applications. As a result, the cost savings, approaches, and efficiencies that virtualization brings allow it to play a crucial role in facilitating innovation and creativity, all leading to creating a market-differentiating vehicle user experience.
Challenge 5: Shortage of Expertise in Software Development and Deployment
With OEMs endeavoring to do more of the SoC hardware and software stack development in-house, a shortage of development expertise in these organizations has been exposed. The SDV, by definition, is causing OEMs to adopt methodologies and techniques that are well-proven in enterprise software development. Adoption of SaaS platforms and the cloud computing are not the only answers.
OEMs do many things extremely well, including cultivating safety culture and certification, hardware/software end-to-end management and integration, and reliability and cost management at volume. Meanwhile, cloud vendors excel at software lifecycle management, elastic scalability to handle dynamic workloads, and ecosystem enablement at scale. However, there is a gap to be bridged in terms of expertise. While critical tools like an embedded RTOS or hypervisor focus on ease of use, and may be “cloudified,” familiarity and expertise with virtualization, containerization, and cloud-native microservices concepts can’t be achieved overnight, nor can agile methodology, CI/CD pipelines, or use of software toolchains.
Bridging this gap requires reliance on a software vendor with a proven, trusted hypervisor and RTOS foundation, as well as cloud experience coupled with in-depth automotive experience and longstanding relationships with the OEMs.
The Road Ahead
Embedded systems virtualization is a critical approach to addressing numerous business challenges today and in the future, as the automotive industry continues to evolve the software-defined vehicle. This evolution is happening not only in the vehicle, but also in the software lab, as OEMs think less about combustion engines and more about software-based connectivity and functionality, and the challenges and opportunities this offers.
BlackBerry® QNX® Hypervisor and QNX® Hypervisor for Safety, allied to our QNX® Neutrino® real-time operating system, comprises some of the industry’s leading automotive foundational software. I encourage you to learn more, or contact us and take our BlackBerry® QNX® automotive software platform out for a spin.