Deterministic network calculus for multicasting : a numerical comparison between explicit intermediate bounds and multicast feed forward analysis
Bruno Oliveira Cattelan · Lume (Universidade Federal do Rio Grande do Sul) · 2018
Nowadays networked systems have become widely spread. Computer networks are not only important for businesses and entertainment, but also for specific safety-critical applications. One important example of a special-purpose network for such safety-critical applications are the Avionics Full-Duplex Switched Ethernet (AFDX) networks, which have been patented by Airbus. These networks require guarantees about their performance, and Deterministic Network Calculus (DNC) has been used to certificate them. An important characteristic is that their data flows are defined as Virtual Links (VL), which can be multicast. However, for a long time DNC was not able to properly analyse such flows. This limitation was previously circumvented by making overly pessimistic assumptions about the demands of the flows in the network. Although valid, this kind of analysis leads to over-provisioning of the network, that in turn means unnecessary increases in cost. In this work, we discuss the two most performant DNC multicast analysis methods presented in the literature. These are the Explicit Intermediate Bounds (EIB) and the Multicast Feed Forward Analysis (MFF). We compare both algorithms to the Unicast Transformation (UT) technique to analyse multicast flows regarding the quality of the bounded delay. We also compare both algorithms to each other, and offer an comparative analysis of their performances.