Adaptive Scheduling for TCP-Fairness in Wireless Broadband Networks

Introduction

Internet is a huge ensemble of highly distributed networks where most of the communication is carried over Transmission Control Protocol (TCP), an elastic transport protocol. TCP provides reliability along with flow and congestion control between the endpoints. User Datagram Protocol (UDP) is another transport protocol which is inelastic in nature and enjoys a small share when compared with TCP in today’s Internet. UDP is mainly used for real time communications which is delay sensitive (Ex. A VoIP phone call).

Due to the advances made in multimedia and infotainment services over the Internet, use of UDP is seeing a meteoric rise (The Cooperative Association for Internet Data Analysis, 2010). If we assume that the total bandwidth available per base station (BS) is a constant value (e.g., say 100 Mbits) and the bursty UDP traffic increases, then the existing TCP traffic will get squeezed and the user experience for applications using TCP will get deteriorated. It is interesting to note that even few of the real time services, (Ex. Skype) use TCP for its transport (Baset & Schulzrinne, 2006). They will also incur a severe degradation in voice/video quality due to increasing UDP flows, which is bursty in nature. The negative impact of the increasing UDP traffic in Internet ranges from unfairness against competing best effort TCP traffic to the potential for congestion collapse which severely impacts the responsiveness of TCP based services (Floyd & Fall, 1999). UDP traffic also suffers with jitter by TCP’s continuously changing window size. However, in this chapter we specifically focus on the challenges in protecting TCP throughput in presence of UDP flows by means of isolation of TCP and UDP traffic.

In the past decade, the problem of TCP protection has been attracting attention of researchers around the world and a number of mechanisms/architectures have been suggested on the Internet core and edge routers to guarantee a minimum performance to the TCP services. A number of router mechanisms have been suggested with an essential goal of end-to-end congestion control that includes identification and restriction of bandwidth hungry flows that are unresponsive and competing with TCP. But unfortunately, this problem is not fully solved in access network which connects the end user with the Internet backbone. The problem is common on both wired and wireless, but more acute in wireless because:

• 1.

  Wireless medium is prone to have more losses when compared with wired medium.

• 2.

  Wireless medium has a limited shared bandwidth where wired medium has the luxury to increase bandwidth with increased hardware installations.