An enforced inter-admission delay performance-driven connection admission control algorithm

Abstract

Connection Admission Control (CAC) is an important function in a computer network that supports Quality of Service (QoS). The function of CAC is to decide whether a new connection can be admitted on the network or a part of the network in such a way that the QoS of the new connection and the already established connections will remain within the requested limits. CAC must also ensure that network resources are used efficiently avoiding unnecessary rejections of candidate connections. Some CAC algorithms attempt to estimate or derive available resources in each hop between the source and the destination. Then they use parameters of the candidate flow and the knowledge about the available resources to make the admission decision. Very often such algorithms use measurements to find out availability of the resources. Hence the name of such algorithms, Measurement-Based CAC (MBCAC). Another class of CAC algorithms assumes that it is the end nodes and not the network that should perform the connection admission. These algorithms use probing packets to probe the path between the end nodes for the requested level of QoS. In this paper a heuristic-based per-hop CAC algorithm is introduced that adapts the average rate of connection admission in response to the measured system performance. In particular, the algorithm decreases the rate of the connection admission when the system is overloaded and increases it when the system is underutilized. The paper uses simulations to show that the algorithm is free from the shortcomings of the current MBCAC algorithms.