Joint Selection Of Source Class And Call Admission Policy For The Bounded Delay Services

Abstract

An essential challenge of incorporating VBR video traffic into networks with service guarantees lies in the difficulty of finding an appropriate traffic characterization that captures the dynamics of the source. A simple traffic model, such as leaky bucket, will lead to low network utilization, especially for burst traffic. However, both multiple leaky bucket and D-BIND are more accurate traffic model, but it is more complex to implement the policy function for high-speed network. And also close form formula of the call admission algorithm is not easily to be derived. In this study, we improve the network utilization by using s(r) (burstiness curve) traffic specification instead of single set of (s,r) while the complexity of the network still same as the network of single leaky bucket traffic model. We derived the close form formula of call admission control and show how to select the optimal class from the burstiness curve . After the call admission algorithm gets the optimal class, it polices the input traffic with this set of (s,r) parameters by using a leaky bucket regulator only the call admission algorithms are different. Therefore, the regulator is only a leaky bucket at first node that the connection passes through. The network is same as single leaky bucket traffic model. The performance is over 1.21 times for burstiness curve model to single (s,r) model in our experiments.