Optimal routing and scheduling of complemental flows in converged networks

Abstract

Converged networks support applications with completely different (real-time) requirements. The communication paradigms offered in converged networks are predominantly treated as separate entities from the perspective of traffic engineering, e.g., time-triggered traffic for closed-loop control systems, shaped traffic for multimediastreaming applications, and best-effort traffic for non-time-critical IT applications. However, there are scenarios where applications benefit from considering time-triggered messages and non-timetriggered messages as complemental components of a single traffic flow. These applications have the property that time-triggered transmissions guarantee basic functionality (e.g., stability of a control system), and additional non-time-triggered transmissions improve the application's performance. We present how to model these so-called complemental traffic flows for this type of application using a traffic metric for the description of the non-time-triggered traffic part. Furthermore, we show that complemental flows are suitable for traffic engineering by presenting two different approaches for the problem of optimized joint routing and scheduling in converged networks with mixed integer linear programming. In our evaluations, we use an exemplary min-max objective for the joint routing and scheduling problem which yields an average reduction of the peak value of the traffic metric by 20-30% over constraint-based approaches. CCS CONCEPTS • Networks → Network algorithms; • Computer systems organization → Real-time systems.

Dynamic QoS-Aware Traffic Planning for Time-Triggered Flows in the Real-Time Data Plane

“…runtime and schedulability. For example, for meshed networks, it is intuitive to expect better results with more candidate paths at the price of longer runtimes [37], but we expect a diminishing return with regard to the number of candidate paths due to shared sub-paths on the k-shortest paths. Therefore, we use 3-5 candidate paths per flow, which already yields a very high schedulability in our evaluations (similar to [16]).…”

Time-Triggered Traffic Planning for Data Networks with Conflict Graphs

“…Additionally, we also implemented an adapted version of a recently proposed, typical ILP [14] for the traffic planning problem to compare the conflict-graph based traffic planning approach to a constraint-based approach which operates directly in the domain of network and streams. We adapted the ILP to use the same system model of the traffic planning problem as CGTP, i.e., multiple candidate paths are considered, and we use a time-discretization to model temporal constraints.…”

“…The constraint-based programming approaches either consider only the scheduling aspect [10], [11], or scheduling and routing [12]- [14]. In these approaches, a set of constraints is formulated using the respective framework (satisfiability modulo theories, array coding, integer linear programming) in the domain of the network problem, e.g., constraints directly restrict the transmission schedule or the paths of packets.…”