Since various kinds of applications in communication networks have emerged, many researchers have considered flow control algorithms based on utility functions characterizing them. When it comes to fairness issue, however, those algorithms still deal with it in bandwidth point of view, not in utility point of view. This is mainly because those algorithms are based on the optimization frameworks which inevitably impose some restrictions on utility functions.
In this paper, we develop a simple method achieving fairness criterion so called Utility Max-Min Fairness. Actually, it uses a explicit control protocol called XCP and is not based on optimization frameworks. Since it modify only end-user part of XCP, it is very simple and quite scalable if we assume that XCP router are deployed.
In addition, we define a Threshold Problem which addresses a situation where a certain utility threshold can not be satisfied even under utility max-min framework. We develop a method to deal with this problem, performance of which depends on statistical time division multiplexing.
Packet-level simulations show that those methods are working properly.
이 논문에서는 기존의 만족도를 최대화 시키는 최적화 문제 형태의 흐름제어 기법을 탈피하여 XCP를 이용한 흐름제어 기법을 제안한다. 제안하는 흐름제어 기법은 기존의 최대-최소 공평성의 일반화된 형태인 만족도 최대-최소 공평성을 달성한다. 또한 제안한 방법으로도 만족도의 문턱값을 만족시키지 못할 경우에 대한 해결방법을 제시한다. 시뮬레이션 결과는 제안한 방법이 적절히 동작함을 보여준다.