Date Added: Jan 2011
The authors derive the maximum entropy of a flow (information utility) which conforms to traffic constraints imposed by a generalized token bucket regulator, by taking into account the side information present in the randomness of packet lengths. Under constraints of maximum aggregate tokens and maximum aggregate bucket depth, information utility is maximized only if the generalized token bucket regulator is a standard token bucket regulator. However, if state-based choice of regulator parameters is allowed, then the generalized token bucket regulator can achieve a higher information utility. This may have an impact on a pricing policy based on regulator parameters.