The evolution of Software Defined Vehicles (SDV) has led to increased performance requirements for communication middleware, particularly to support continuous feature updates across both current and future platforms. To address these requirements, the architecture of communication middleware must be tightly integrated with the underlying operating system and adhere to well-defined design patterns to ensure reliability, scalability, and real-time performance. In traditional vehicle architectures, the electrical/electronic (E/E) architecture is typically fixed for each vehicle generation. However, in the context of Software-Defined Vehicles, many software capabilities are increasingly decoupled from the underlying E/E architecture. As a result, network efficiency becomes critical to enabling the extension and upgrade of software functionalities, particularly through over-the-air updates, on existing ECU platforms. The Scalable Service-Oriented Middleware over IP (SOME/IP) protocol often serves as the backbone for vehicle Ethernet communication. We evaluated a version of the SOME/IP stack directly integrated into the network stack of QNX operating system, and compared it to the traditional approach, where a daemon is running as a standalone process. Additionally, the stack was optimized to support a scalable multithreading paradigm and low-level IPC paths between involved processes were optimized to work more efficiently with the underlying SHM and notification primitives. For meaningful results, we simulated a real-world use case with realistic vehicle communication. Our measurements revealed a substantial reduction in CPU usage by 50.2% and a very low latency of 55 µs. The direct integration of the network stack eliminates unnecessary context switches, memory copies, and expensive IPC communication between a daemon and the network stack, while the optimizations to threading and IPC mechanisms target the application side, significantly enhancing performance. The tight integration of the SOME/IP functionality into the QNX network stack with the “Foundational Vehicle Software Platform” (called “Alloy Kore”) of QNX and Vector represents an effective strategy to improve middleware performance. With a multi-threaded architecture, the communication middleware can manage significantly higher volumes of network traffic while maintaining a lower impact on CPU utilization, exploiting the parallel processing power of modern hardware.
Maidl et al. (Tue,) studied this question.