Browsing by Author "Dr. William Stewart, Committee Member"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • No Thumbnail Available
    Analytical Models and Efficient Dimensioning Algorithms for Communication Systems In Randomly Changing Traffic Environments
    (2009-01-25) Tian, Wenhong; Dr. Khalid Harfoush, Committee Member; Dr. William Stewart, Committee Member; Dr. Harry Perros, Committee Chair; Dr. Mike Devetsikiotis, Committee Member
    Analysis and design of communication systems in today's dynamically changing environment face a lot of challenges. Communication systems are not static but are dynamically changing today. Networks can become overloaded at one moment and underutilized at the next. Humans simply cannot respond to changes fast enough; computer programs must be used. Queueing theory has typically been used to address queueing problems of communication networks under the assumption that the inter-arrival times and the service times are renewal processes, and the arrival rate and bandwidth requirement of each customer is constant. In today's dynamically changing environment, such as in high-speed integrated-services wired and wireless networks and Web service, these assumptions are frequently unjustified. Our research focuses on the performance evaluation and efficient resource dimensioning algorithms of communication systems in a dynamically changing environment. The topics investigated are either new or they have not been studied well in the literature. One of our contributions is providing efficient dimensioning algorithms in closed-form with a constant or near linear complexity. In this thesis, we considered the following cases: a) the arrival rates of customers are time-varying, b) the bandwidth requirement of a customer is variable, c) integrated streaming and data traffic where the available bandwidth for the data traffic is dynamically changing and d) a statistical multiplexer with autocorrelated and bursty arrival processes. This work starts with the dynamic analysis and design of a communication link when the arrivals are time-varying. The time-varying arrivals capture the nature of a dynamic traffic environment. Using transient analysis, we dimension a link under time-varying arrivals and transient load. We compare some methods (both numerical and closed-form) in the literature and propose efficient ways to design the system under study. We next consider the case when the customer's bandwidth requirements are time-varying. A multi-rate loss model with class-change is proposed and the blocking probability is calculated using an iterative recursive formula. The optimal dimensioning of the capacity of this system is obtained using a formula with a constant cost. Subsequently we consider a model to analyze the integrated streaming traffic and data traffic, where the data traffic is bursty with a time-varying bandwidth shared with the streaming traffic. Using priority decomposition, we proposed closed-form solution for dimensioning of both streaming and data traffic. Finally we analyze a statistical multiplexer in a packet-switched network where the arrival process is the superposition of n independent identical bursty and autocorrelated arrival processes. We construct the superposition of these arrival processes and characterize the departure process from the multiplexer by exact Laplace transform for the first time and a two-stage Markov-modulated Poisson process. These results permit us to analyze and dimension the system more efficiently under study.
  • No Thumbnail Available
    Performance Analysis of Optical Burst Switched Networks with Dynamic Simultaneous Link Possession
    (2005-11-21) Battestilli, Tzvetelina; Dr. Arne Nilsson, Committee Member; Dr. Yannis Viniotis, Committee Member; Dr. Harry Perros, Committee Chair; Dr. William Stewart, Committee Member
    Given the current state of the technology, the Optical Burst Switched (OBS) architecture is a practical optical switching solution for the optical networks. In OBS, the user data is transmitted in variable size data units, called bursts, which travel as an optical signal along the entire route. The control information for each burst is transmitted prior to its corresponding burst and it is electronically processed at each hop along the route. The dynamic nature of OBS allows for network adaptability and scalability, which makes it very suitable for the transmission of bursty traffic. In this thesis we study and analyze the performance of OBS networks. We consider the case when the bursts are large enough to simultaneously hold wavelengths on multiple links along the route. Since the size of the bursts varies and the link distance between two adjacent network nodes also varies, a burst may simultaneously occupy wavelengths on a variable number of links as it travels from its source to its destination. As the burst propagates through the network, it dynamically acquires and releases wavelengths from link to link. In this thesis, we propose queueing network models that feature dynamic simultaneous link possession and analyze them in order to obtain the end-to-end burst loss probabilities. This thesis is structured into five parts. In the first part, we describe the Optical Burst Switch architecture and summarize its main design components. In the second part, we define and motivate the simultaneous link possession problem. In the third part, we study an OBS network where the bursts are large enough to simultaneously hold wavelengths on one or two consecutive links. In the fourth part of this thesis, we analytically study an OBS network with simultaneous link possession and a very large number of wavelengths per link. In the fifth part, we address the problem of how to analytically evaluate the performance of OBS networks when the arrival traffic is bursty.