A congestion control algorithm for max-min resource allocation and bounded queue sizes

This paper deals with the congestion control problem in computer networks which is viewed as a resource allocation problem constrained by the additional requirement that the queue sizes need to be bounded. We propose a distributed algorithm which converges to the max-min fair allocation of resources among the users of the network and at the same time ensures that the buffers are either empty or track a reference queue size. The problem is formulated mathematically and the proposed algorithm is shown analytically to fulfil the design objectives. The local asymptotic stability of the equilibrium point is established. The problem can be viewed as a hybrid system with changing affine dynamics in different regions of the state space. The transient performance of the proposed algorithm is evaluated through simulations using Matlab. The algorithm can form the basis for the development of an end-to-end communication protocol since it requires no maintenance of per flow states within the network.