RUB » Bibliotheksportal

Alexander Scheffler, Jens B. Schmitt, Steffen Bondorf

• Network technologies are being developed to increase not only performance of data communication but also for provision of deterministic guarantees. Such designs foster the development of distributed real-time applications. When networks must provide time-sensitive communication, a commonly found design feature is First-In First-Out (FIFO) – a natural design for multiplexing different data flows into queues and for scheduling queued data. Alongside technological advancements, there is the development of formal tools to reason about timing behavior. Network Calculus is such a methodology. It has been widely adopted already and its modeling capabilities were extended to features found in modern standards as, e.g., in IEEE Time-Sensitive Networking. However, the basic challenge to compute a deterministic bound on a flow’s worst-case end-to-end delay in FIFO networks still imposes challenges. Different analyses exist but often offer limited scalability. In this paper, we present an analysis with a readily tunable tradeoff between quality of the delay bound and the computational effort it imposes. We combine a greedy algorithm with an iterative directed search whose termination criterion can be adapted, e.g., subject to the analyzed network size. Our approach provides bounds of competitive quality at smaller computational cost than current alternatives for the analysis of feedforward FIFO networks.