SF-RED - a novel server-based AQM to provide inter-server fairness service

Abstract

We design an efficient active queue management (AQM) algorithm, SF-RED, that aims to provide inter-server fairness service in a simple and scalable manner. In the TCP/IP network, it happens that one may intentionally, but in a legal way, generate many "flows" for the same job in order to grab more bandwidth. This leads to servers' bandwidth allocation being unfair, and thus makes ISP price-based services policy enforcement difficult. We define this as an inter-server fairness problem. There were more than 50 new AQM algorithms which have been proposed since 1999. However, none of them consider the inter-server fairness problem. We study the inter-server fairness problem in traditional AQM, and propose a novel server-based AQM algorithm for providing fair services to different servers without affecting congestion control performance. We prove the effectiveness of our approach with extensive simulations using ns-2 simulator.