An Empirical Study of Tightest Network Calculus Analyses for Networks with Multicast Flows

The Network Calculus (NC) analysis is concerned with deriving an upper bound on the end-to-end delay of data flows. For tighter bounds, the analysis needs to incorporate features of the network model in the best way possible. This has proven to be a non-trivial task, resulting in what seemed a succession of improved NC capabilities. However, it was discovered that neither of the two most prominent analyses in the literature is always best. Deriving the tightest delay bound thus became an expensive search for the most beneficial tradeoff between drawbacks of alternative NC analyses. In networks with multicast flows, this problem is amplified, yet, while causes for this situation are known, there is no empirical investigation on the ratio between different analyses obtaining the tightest bound. In this paper, we provide an empirical study of the distribution of delay bounds derived with NC analyses that can be used to bound delays in networks with multicast flows. We do so by evaluating differently sized avionics-like networks.