Toward a Resilient Automotive Service-Oriented Architecture by using Dynamic Orchestration

Modern software development in vehicles is focusing on a service-oriented approach. Structuring software systems into self-sufficient software components that provide specific capabilities to the overall system allow software engineers to make changes to vehicle functions more granularly. The decentralized SOA approach offers advantages, as it enables loose coupling between components instead of statically implementing their relationships. But with the increasing degree of autonomy and dynamism of the vehicle's software, the system's safety and security requirements are also growing. Preventive measures will no longer suffice here; instead, resilient systems are required that provide a minimum level of safety even in the event of an unexpected problem. Today, a SOA's services are assigned to a hardware platform during development and executed there, which lacks being able to react to problems or changing requirements. One possibility for being more flexible at runtime, is the use of an orchestrator, which dynamically allocates resources to services while retaining the advantages of a loosely coupled architecture. This paper proposes a methodology for implementing a resilient vehicular electronic architecture based on orchestrating containerized software. To avoid a single point of failure, a distributed approach for a dynamic orchestrator that deploys the software to appropriate execution platforms is proposed. The orchestrator makes its deployment decisions based on specifiable parameters (e.g., required RAM, GPU) and dependencies between services. The decision process adapts to changes in these factors dynamically, making the system able to react to external influences. The concept differentiates itself from other approaches by tracking dynamic changes to specified parameters and easily extensible interfaces for new parameters or requirements. In addition, the concept introduces a priority metric to describe the impact of services in the system and models how this metric is inherited through dependencies. The concept is evaluated qualitatively by three exemplary use cases, demonstrating the effect of dynamic orchestration on the resilience of the vehicle.