Using Multi-Link Grouping Technique to achieve tight latency in network calculus

Abstract

It is of great importance and a challenge to compute tighter upper bounds on the latency for flows in the avionics distributed real-time system with strict time constraint. A neglect of flows serialization on the link is one of the significant reasons for the untight latency estimation. This paper proposes a multi-link grouping technique in the Network Calculus to achieve tighter worst-case end-to-end latency for flows in the AFDX (Avionics Full Duplex Switched Ethernet) network, with FIFO (First In First Out) and SP (Static Priority) assumptions. In addition, we demonstrate that such technique will only have influence on arrival curve but nothing to do with service curve with FIFO and SP scheduling policy, even though SP scheduling is closely associated with arriving characteristics of flows. In the end, tighter response latency bounds can be obviously observed in the comparison experiment with and without multilink grouping technique.