Dynamic Resource Allocation Based on Weighted Max-Min Fairness

Abstract

In this paper we extend weighted max-min fairness by adding strategies that enable users to modify their fair allocation by sharing and borrowing bandwidth with other users. There is also a mechanism to reclaim bandwidth a user has previously shared. The main focus of our investigation is to develop a method for allocating link bandwidth between multiple users on multiple links of a communication network in the face of changing demands, usage patterns, and priorities while still being guided by max-min allocation. The algorithm is generic and could be applied to settings other than data networks such as process scheduling in operating systems, etc. We have tested the feasibility and performance of the algorithm via simulation. We have also integrated it into a commercial product and tested it in a real-world setting. We present the results from both types of tests in this paper.