ATM(Asynchronous Transfer Mode) network can support various QoS (Quality of Service). ATM Forum defined 5 service categories with regard to traffic characteristics and QoS requirement. ABR is one of them, and its objective is to support non-real time data traffic. Nevertheless, if we use ABR for video traffic, we can combine the desirable features of other services: simple admission control; frequent bandwidth negotiation, etc.
There are many criteria for ABR rate allocation algorithms. Fairness is one of the most important criteria. We focus on fairness for video services. The goal of Max-Min Fairness (MMF), the most widely used fairness criterion, is to achieve the fairness of source rates, i.e., equal bandwidth sharing. But, MMF is inappropriate for video services, which require different amount of service rates for the fairness of video quality. So, a WMMF(Weighted MMF) scheme which uses the varying rate demands of sources as weight, was proposed. But, using that scheme combined with existing rate allocation algorithm, we can't obtain fairness between video sources, because the convergence of rates to fair rates is not fast enough.
In this thesis, we restrict the environment to LAN in which RTT(Round Trip Time) is small, and is not fluctuating a lot. In such environment, we support fairness between video sources by deciding the source rates after the values of ER feedback converges to the fair rates and by the earlier prediction of data rates. And, we also mention the relation between parameters not to go against fairness.
