Tradeoff between the cycle complexity and the fairness of ring networks

In this paper we study performance tradeoffs of fairness algorithms for rin g networks with spatial bandwidth reuse, by using two measures: (i) the fai rness cycle size as a complexity measure; and (ii) the Max-Min optimal fair ness criterion as a throughput measure. The fairness cycle size is determin ed by the number of communication links involved in every instance of the f airness algorithm (several identical fairness algorithms can be executed co ncurrently on the same ring). The study compares three fairness algorithms with different cycle sizes: the Global-cycle algorithm (implemented in the Serial Storage Architecture - SSA) in which the cycle size is equal to the number of links N in the ring; the Variable-cycle algorithm in which its cy cle size changes between I and N links; the One-cycle, where there is a fai rness cycle on every Link. How the varying cycle size affects the network p erformance with respect to the Max-Min optimal fairness criterion is shown. The results show that for non-homogeneous traffic patterns, decreasing the fairness cycle size, while increasing the complexity, can significantly im prove the performance with respect to the Max-Min optimal fairness criterio n. (C) 2001 Elsevier Science B.V. All rights reserved.