Implementation and Evaluation of Differentiated Services on Linux

Abstract

The existing Internet architecture is based on the “best effort” model for delivering packets across the Internet. The current architecture delivers a packet at its best possible (best-effort) but doesn’t guarantee when it will be delivered. Nowadays, the users work and play habits are changing, e.g. users expect to watch movies through the network, play 3-D games, and check their stock online, videoconference and other. The demands of the users have changed dramatically since the creation of IP, where it was mostly used for email and ftp. Another new application is the WWW that has been widely used worldwide. WWW has created a new friendly interface for the user, and stimulated further demands from the network. The existing architecture of IP is inadequate to handle new applications. Time critical applications such as video, audio and several others have created an even greater demand on the Internet. Recently, several different solutions were proposed, but most have failed to replace IP. Lately, several new protocols and architecture were proposed to enable basic quality of service provision in Internet. In this thesis we investigate the Differentiated Services (DiffServ) architecture. DiffServ is a new architecture based on the concept of aggregated differentiated treatment of services. DiffServ was proposed in 1997. Since then it has attracted a lot of attention by many researchers. The aim of this thesis is to implement a differentiated services pilot network in Linux environment and investigate the performance of various network functions, that may provide differentiated quality of service. These functions include various queuing disciplines such as pFifo, RED, and TBF. Through the pilot network we aim to investigate different ways to implement differentiated networks and present recommendations for different network traffic and conditions.