AI-based mission-critical software exposes a blessing and a curse: its inherent statistical nature allows for flexibility in result quality, yet the mission-critical importance demands adherence to stringent constraints such as execution deadlines. This creates a space for trade-offs between the Quality of Result (QoR)-a metric that quantifies the quality of a computational outcome-and other application attributes like execution time and energy, particularly in real-time scenarios. Fluctuating resource constraints, such as data transfer to a remote server over unstable network connections, are prevalent in mobile and edge computing environments-encompassing use cases like Vehicle-to-Everything, drone swarms, or social-VR scenarios. We introduce a novel approach that enables software engineers to easily specify alternative AI service chains-sequences of AI services encapsulated in microservices aiming to achieve a predefined goal-with varying QoR and resource requirements. Our methodology facilitates dynamic optimization at runtime, which is automatically driven by the MARQ framework. Our evaluations show that MARQ can be used effectively for the dynamic selection of AI service chains in real-time while maintaining the required application constraints of mission-critical AI software. Notably, our approach achieves a 100x acceleration in service chain selection and an average 10% improvement in QoR compared to existing methods.

At first glance, running mission-critical applications at the edge appears to be an opportunity to benefit from scalability and reusability. The low latency to the edge makes it particularly interesting for mission-critical applications. The hardware heterogeneity of the edge, coupled with the strict requirement for the execution time of a mission-critical application, creates the need for flexible application control and, at the same time, increases the complexity of modeling such systems. With its Feature Models (FMs), software product line engineering offers a modeling option for various alternative compositions of an application. However, the calculation of valid configurations takes too long for the dynamic adaptation of an application flow of a mission-critical application. This paper presents an approach for slicing FMs to support mission-critical applications. Our approach supports the strict requirements on the execution time of mission-critical applications.

Time-critical requests are becoming increasingly important for microservice service chains. For such service chains, a late response may be worthless or cause failures. Previous works proposed to prevent late responses by trading computation time for result quality, depending on the available resources. However, determining the Quality of Result (QoR) for each operation in a service chain ahead of time, as is the state of the art, cannot prevent requests that are currently processing from being late if resource availability changes, e.g., due to resource sharing at inelastic Edge sites. Therefore, we present a framework to control the QoR online by replanning QoR values whenever resource changes are detected.

Mission-critical applications, which must adhere to processing deadlines, can benefit from low latencies offered by the edge. Adapting the Quality of Result allows for targeted processing times by selecting various approximations. Feature models can be employed to manage the resulting multitude of possible configurations. However, the deployment and (re)-configuration process is very time-consuming, making it impractical for mission-critical applications. In this work, we introduce a processing pipeline with components that significantly accelerate online (re)-configuration based on changing latencies compared to the state-of-the-art. Additionally, we address the edge-specific discovery of potential microservice chains capable of executing the application.