Stochastic evaluation of fair scheduling with applications to quality-of-service in broadband wireless access networks

• Authors: Mohammed Hawa;David W. Petr
• Venue: Stochastic evaluation of fair scheduling with applications to quality-of-service in broadband wireless access networks
• Year: 2003

A prerequisite to providing multimedia services over wireless packet-switched networks is the provisioning of new mechanisms that allow such networks to treat packets differently according to their limitations and performance bounds. Those new mechanisms are collectively included within what is called a Quality-of-Service (QoS) architecture. A desirable feature of a QoS architecture for wireless networks is the ability to provide wireless end users with the same QoS guarantees that wired users currently enjoy. To that end, we develop a new and efficient scheduling architecture to support bandwidth and delay QoS guarantees for packet-switched Broadband Wireless Access (BWA) networks. Our design goals are simplicity and improved network performance. The architecture we develop in this research effort supports various types of traffic including constant bit rate, variable bit rate (real-time and non-real-time) and best effort. A key element of our proposed QoS architecture is known as Fair scheduling or Fair Queueing (FQ). Fair scheduling algorithms have received much attention in recent years because of their ability to provide a wide range of QoS guarantees to end users. In this dissertation, we concentrate our efforts on analyzing the stochastic performance of such a class of fair scheduling systems. We start by presenting a new analysis method that results in reasonably tight upper and lower bounds on mean packet delay and mean buffer occupancy experienced by fair scheduling algorithms under Poisson arrivals. We coin the new term M/G/FQ to describe this analysis method, and provide a range of simulation experiments to validate its results. Using more simulations, we continue our study of fair scheduling algorithms by comparing the performance of three packet-based FQ policies to each other and to the reference scheduling policy called Generalized Processor Sharing (GPS) under random Poisson arrivals. Our experiments allow us to derive many useful insights into the operation of such packet-based FQ policies, which helps us better understand their specific characteristics and properties.