In this paper, we explore the potential of applying sojourn-time analysis in QoS metrics specification and in CAC algorithm design for ATM networks. We propose using the probability a link overflow occurs during a connection’s life as a measure of its QoS. Such sojourn-time based QoS metrics have big impacts on many network traffic management components, eg., CAC and routing. A CAC policy is accordingly designed which guarantees the so specified QoS for the entire life of each connection in the system. For a basic model, simple formulas are obtained for the estimate of the probability that a link overflow occurs within certain time interval. The analytical approach used and presented in the paper is a set of general results for estimating the probability distribution function and the mean of the sojourn-time of a Markov process being in a subset of its state space, which may correspond to a system’s underload status.
[1]
K. Qian,et al.
Capacity of ATM Switches
,
1993
.
[2]
D. Mcdonald,et al.
Asymptotics of Exit Times for Markov Jump Processes I
,
1994
.
[3]
Asymptotics of the Exit Distribution for Markov Jump Processes; Application to Atm
,
1994,
Canadian Journal of Mathematics.
[4]
J. W. Roberts,et al.
Performance evaluation and design of multiservice networks
,
1992
.
[5]
San-qi Li,et al.
Sojourn-time analysis on nodal congestion in broadband networks and its impact on QoS specifications
,
1996,
Proceedings of IEEE INFOCOM '96. Conference on Computer Communications.
[6]
Thomas M. Liggett,et al.
Exponential $L_2$ Convergence of Attractive Reversible Nearest Particle Systems
,
1989
.