Correlated token bucket shapers for multiple traffic classes

Abstract

Traffic shaping is an important technique for quality of service guarantee in a network that provides integrated services. It Is often used in an access node to regulate user traffic submitted to the network. Since the leaky bucket algorithm has been selected by international standard organizations as the traffic policing algorithm, an ideal shaper should change the temporal characteristics of user traffic to conform to the criterion. It is well known that token bucket is equivalent to leaky bucket and is often used to name shapers that shape traffic in accordance with the leaky bucket criterion. However, in standard documents, the leaky bucket algorithm was described for a single traffic class. There is no discussion for multiple traffic classes. One possible way is to use multiple independent shapers such that each traffic class is regulated by one shaper independent of other traffic classes. This approach is simple but may not be reasonable. In this paper, we propose to use multiple correlated token buckets to shape user traffic with multiple classes. In our proposed architecture, cheaper traffic classes can utilize the unused bandwidth subscribed for more expensive traffic classes. Besides, each traffic class Is guaranteed a minimum bandwidth specified by its leaky bucket parameters.