Browsing by Author "George N Rouskas, Committee Chair"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • No Thumbnail Available
    Congestion Control and Quality-of-Service (QoS) on Jumpstart Optical Burst Switched Environment
    (2006-07-16) Yang, Li; Wenye Wang, Committee Member; Khaled Harfoush, Committee Member; Rudra Dutta, Committee Member; George N Rouskas, Committee Chair
    This thesis studies the congestion control and Quality-of-Service (QoS) problems in Optical Burst Switched (OBS) networks. It consists of three parts. In the first part, we consider path switching as a congestion control mechanism at the edge of the network. We study a suite of path-switching strategies, each of which gives a different method to estimate the path congestion online. We also develop a framework for combining several path switching strategies into hybrid strategies whose results are based on the decisions of multiple individual methods. We demonstrate the effectiveness and benefits of adaptive path selection via simulation. In the second part of the thesis, we develop a general framework for absolute service guarantees for an OBS network in terms of the end-to-end burst loss. We first present a parameterized model for wavelength sharing. Then, we develop a heuristic for optimizing the policy parameters to support per-link absolute QoS guarantees. Finally, we present a methodology for acquiring the per-link parameters from the end-to-end QoS requirements so as to provide network-wide guarantees. We present numerical results to validate our approach. In the third part, we present a per-link wavelength provisioning scheme based on Constrained Markov Decision Processes (CMDP) theory to provide service differentiation. Service differentiation is evaluated with two objectives on OBS networks: to maximize the constrained throughput; and to minimize the loss of the best effort traffic subject to the constraints on the priority traffics. The randomized threshold policies we obtain are simple to implement and operate, and make effective use of statistical multiplexing.
  • No Thumbnail Available
    Economic Models for Tiered Network Services
    (2010-04-12) Lv, Qian; Wenye Wang, Committee Member; Rudra Dutta, Committee Member; Khaled Harfoush, Committee Member; George N Rouskas, Committee Chair
    Tiered network services have been prevalent in current networking industry. The term tiered service means that the network operator only provides a small set of tiers (levels) - which offer progressively higher levels of service - to the customers each of whom will be mapped to one of the given tiers. In this thesis we focus on the economic issues for tiered network service. We first formulate the problem of selecting service tiers from three perspectives: one that considers the users interests only, one that considers only the service provider’s interests, and one that considers both simultaneously. We consider the solution to this problem under two cases: 1) the discrete case, i.e., each user demand is known to the service provider; 2) the stochastic case, i.e., the service provider only knows the probability distribution of the user demands. For both cases, we present accurate and efficient algorithms based on dynamic programming. After finding the set of (near-) optimal service tiers, we then employ game-theoretic techniques to find an optimal price for each service tier that strikes a balance between the conflicting objectives of users and service provider. This work provides a theoretical framework for reasoning about and pricing Internet tiered services, as well as a practical toolset for network providers to develop customized menus of service offerings. We further consider some advanced economic topics in tiered network service. We notice that some network services may tend to be elastic, i.e., the users may value a given service differently and show different willingness to pay for the service. In this thesis, we assume that users are partitioned into some distinct classes. We develop an optimal algorithm to select jointly the set of service tiers and their prices so as to maximize the provider profits. Our research shows that introducing multiple tiers can be an effective market segmentation strategy that may lead to an increase in profits. Another advanced topic in tiered network service is service bundling, which means the network providers combine several services together and sell them as a single package at a lower price than that if the services are sold separately. Based on tools from microeconomics and utility theory, we developed an efficient method to find tiered structures for bundles of network services with the objective of maximizing provider profits under user constraints.