Providing flow-based quality-of-service control in a large-scale network

Abstract

QoS (Quality-of-Services) technologies are very important for services like video streaming and real-time communications because they require continuous fixed bandwidth. In this paper we propose a new "feedback and distribution method" that provides per-flow based QoS for large-scale networks. We enhanced the method's ability to control jitter when distributing streaming data to a large number of users. TCP traffic is burst-type traffic and is used for Web navigation and E-mail, and UDP traffic is constant streaming traffic used for real-time applications. A new method to provide ene-to-end bandwidth control for all users in a large-scale network where streaming services are expected to be very important commercially. The proposed method enabled streaming stable distribution because streaming packets that were susceptible to jitter were protected from burst-data by dividing the input for each different type of flow. Traffic bursts absorbed by setting the traffic measurement interval to an optimal length. Our simulation results showed that streaming data with high bit rates hardly affected TCP data. Therefore, the proposed method stably distributed streaming data, and furthermore, it satisfactorily managed a mixture of various types of data.