End-to-End Delay Approximation in Packet-Switched Networks (2003.08780v2)

Abstract: In this paper, I develop a generalized method to approximate end-to-end delay (average delay, jitter and density functions) in packet-switched networks (PSNs) of any size under 1) Kleinrock's independence assumption (KIA) and 2) when packet lengths are kept unchanged when they traverse from node to node in a network, which is an Alternative to Kleinrock's independence assumption (AKIA). I introduce a new phase-type distribution $C(\mathbf{p},\boldsymbol \theta)$; and then use results from the network flow theory and queueing theory to show that the end-to-end delay in PSNs under KIA and AKIA are two different random variables approximately described by $C(\mathbf{p},\boldsymbol \theta)$. When PSNs have AKIA, I show from simulation that the method under AKIA significantly reduces end-to-end delay approximation errors and provides close approximation compared with the method under KIA.