Admission control for hard real-time connections in ATM LANs

Static priority scheduling for ATM networks

“…The rate remains constant along the route, but the burstiness increases as a function of the accumulated worst-case delay on the route. This theorem generalizes earlier results by Cruz [4] and Raha et al [19], where the traffic characteristic function at the output of an FCFS server was obtained in terms of that at the immediate input to the server [4] or at the source [19]. In Theorem 1 we extend these results to servers with static-priority scheduling.…”

“…Nevertheless, considerable progress has been made recently towards obtaining delay bounds for network traffic in a variety of settings. Much of the previous studies on meeting end-to-end deadlines in ATM networks have concentrated primarily on the design and the analysis of scheduling policies for ATM switches [3,6,8,11,17,19,25]. The efficient implementation of most of these schemes turns out to be difficult in high-performance switches.…”

“…Traditionally, servers are classified into two categories: constant servers and variable servers [4,18,19]. A constant server delays each cell by a constant amount of time.…”

“…Before the first iteration, vectord 0 is initialized to bed 0 := 1; 1; : : : ; 1 : (18) During the n th iteration, the new value ford is computed as follows:d n :=Zd n,1 : (19) In order to demonstrate the convergence of this iterative procedure, we need to determine the error betweend and d n , the vector at the end of the n th iteration. That is, we need to establish the difference between the value of d p;j computed at the end of the n th iteration and the real value of d p;j .…”

“…Even if the system is stable, an iterative procedure must be used for the delay computation to accommodate these feedback effects. In [19], the stability problem was addressed for systems with FCFS scheduling. Determining worst-case delays with FCFS scheduling in a system with traffic dependencies was dealt with in [14].…”