A Study Of Application Oriented Fairness In Network Bandwidth Allocation

This paper is concerned with bandwidth allocation problems in computer networks in which real-time applications may have hard quality of service (QoS) requirement. Traditional fairness definition and bandwidth allocation approaches are unable to deal with these problems since utility functions generally do not satisfy the strict concavity condition in real-time applications. If we consider different application requirements among network users, it may be undesirable to allocate bandwidth simply according to the traditional max-min fairness or proportional fairness. Instead, a network should have the ability to allocation bandwidth resources to various users, addressing their real utility requirements.With the concept of utility functions, where each source uses a utility function to evaluate the benefit from achieving a transmission rate, we interpret the resource allocation problem as a global optimization problem. The solution to this problem is characterized by a new fairness criterion, utilityαproportional fairness. Our definition of utility a proportional fairness generalizes utility proportional fairness and includes arbitrarily close approximations of utility max-min fairness. We have developed a new distributed flow control algorithm to achieve utilityαproportional fairness in a given network, where the application's utility is only assumed to be continuously increasing over the available bandwidth. We also present a generalization to the usual fairness index, which we call utility fairness index. With the aid of this new fairness index, one is able to compare the fairness among various utility-aware bandwidth allocation schemes. We propose a distributed approach to achieve utility max-min fairness in network bandwidth allocation. We also proposed a new converge faster algorithm to solve the problem that the original algorithm converges slowly.