The Deterministic Network Calculus Analysis: Reliability Insights and Performance Improvements
Alexander Scheffler, Markus Fogen, Steffen Bondorf · 2018
The cost of delay analysis increases fast with size and complexity of a network. Therefore, many recent research efforts in Deterministic Network Calculus (DNC) focused on the tradeoff between accuracy and cost of its delay-bounding analyses. In this paper, we present insights on reliable, reproducible and performance on both branches of DNC, algebraic analysis and optimization, as well as the tools employed by them. We reveal circumstances causing problems for reliability and reproducibility of DNC's optimization analysis and we investigate the potential to improve computational performance of algebraic DNC. To that end, we present theoretical background on the topic of parallelizing the DNC analyses and an implementation in the open-source DiscoDNC tool. With our proposed approach, we achieve a speedup of analysis times of one order of magnitude.