Combined control of routing and flow: a multipath routing approach

In this paper we propose an algorithm for combined control of routing and flow, which is based on fluid flow deterministic network models. This leads to globally optimal solutions, in the sense that it solves the problem of maximizing an aggregate utility when all possible paths from sources to destinations could be made available. Flow control is carried out in an and to end manner with the user deciding how to split the flow among the paths it is using and with new paths being added on a shortest path basis. The algorithm avoids the use of unnecessary paths and is always guaranteed to converge to the optimal solution in spite of the fact that routing decisions are based entirely on congestion. Despite the significant complexity, as a result of the source routing approach and the fact that globally optimal solutions could involve a large number of paths, this algorithm provides a direction in which to investigate means of guaranteeing to get the most out of a network by providing appropriate alternative routes for flow control.