A decentralized model for virtual path capacity allocation

We investigate the problem of virtual path (VP) capacity allocation in a decentralized way. Users (VP controllers), sharing common network resources, are assumed to control their VP capacities according to their performance measures. We formulate a model that differs from previous ones in two points: first, each user has its own performance objective that is simply assumed to be concave and smooth; second, a user's strategy (VP bandwidth reservation) is constrained between two numbers. Adopting a standard game-theoretic formulation, we prove that there exists a unique network operating point-Nash equilibrium. We show several interesting properties of the equilibrium and also look at the convergence of some distributed algorithms. After studying the combined problem of routing and flow control in parallel link networks, we investigate the problem of VP capacity allocation in general topology networks, where fixed routing is assumed. Finally, examples of the model are presented.