MAX-MIN flow control in a programmable router

Abstract

In this paper, we represent two-dimensional problem in the network resource allocation for MAX-MIN flow control in a programmable router and propose an intelligent explicit rate(ER) allocation algorithm which is based on a novel control-theoretic ER allocation algorithm. By the two-dimensional property, the MAX-MIN flow control has three steady state solutions determined by two system capacities and the resource demands of input flows. In each steady state, different method of rate allocation is required for MAX-MIN fairness and the intelligent proposed algorithm achieves the goal through different adaptive operations for each steady state. The proposed algorithm is simple in that the number of operations required to compute it at a node is minimized, scalable in that per-flow operations including per-flow queueing, per-flow accounting, and per-flow processing are virtually removed except only one and simple per-flow state management, stable in that by employing it, the input flow rates and system queues are asymptotically stabilized at a unique equilibrium point where the MAX-MIN fairness and target queue lengthes are achieved, respectively. We verify the intelligent operations and the performance of the proposed algorithm through simulations in the Intel IXP1200 Software Development Environment. We believe that this work represents a guideline to solve two-dimensional problems in MAX-MIN flow control and the proposed algorithm plays a key role in this direction.