UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Utility-based resource and QoS optimization in packet networks Mohamed, Amr Mahmoud


Resource Allocation (RA) is used to organize the usage of network’s physical resources in such a way that guarantees optimal utilization while providing predictive performance to network flows in terms of inter-flow fairness and guaranteed Quality of Service (QoS). One approach to provide RA which is particularly suitable for traffic flows with relaxed QoS requirements (i.e. elastic traffic) is the Utility-based Resource Allocation (URA). URA assigns a utility function to each individual user flow to measure the degree of satisfaction of this user as a result of assigning a specific share of resources. The objective of the URA techniques is then to partition the network resources to take full advantage of them in satisfying the QoS requirements of each user flow while providing fair allocation of resource among users by maximizing the aggregate utility of all flows. The objective of this thesis is to devise new methods for URA in wired and wireless networks to provide fair resource sharing and predictive flow performance in terms of QoS while guaranteeing best resource utilization. In so doing, we propose a comprehensive set of algorithms that can be used to provide resource optimization both on the link level or on the network level. On the link level, the thesis proposes a group of algorithms for calculating the optimal classification for a set of traffic flows with diverse QoS requirements to a link with predetermined service levels or predetermined class weights. These algorithms can be used to efficiently study the effect of selecting service levels or class weights according to the distribution of the QoS requirements of the incoming traffic flows. For links with adjustable service levels, we propose two algorithms (OQP, and OQP-OBA) that calculate the optimal partitioning of traffic flows, the best service levels, and the optimal bandwidth allocation to minimize the quantization overhead as a result of QoS-based partitioning. Our simulation results for both link models show that using 4 or 5 service levels will achieve the trade-off between complexity and service level granularity irrespective of the QoS distributions. These algorithms indeed provide major enhancements in that fairly unexplored area. On the network level, a key contribution of this thesis is the development of a new decentralized algorithm (ORAWM) for resource optimization over multihop wireless networks. The algorithm is used to control the rates of the end-to-end sessions utilizing the bandwidth-efficiency feature of multicast to provide resource optimization in a totally distributed network environment without any synchronization requirements between network node calculations. Through analytical modeling and simulations, we prove the convergence of the asynchronous algorithm under slow network changing conditions such as channel capacity and node mobility. We also devise a detailed network architecture and discuss the protocol implementation for deploying ORAWM in an ad hoc network. We also extend our solution to include multicast sessions with heterogeneous receivers (ORAHWM) and discuss the modified network architecture to support, multirate multicast trees. The results show that ORAHWM not only provides flexibility in allocating resources across multicast sessions, but it also increases the aggregate system utility and improves the overall system throughput by almost 30% compared to homogeneous multicasting (ORAWM). We also provide a comprehensive set of simulations that show the effect of deploying these algorithms on the overall resource utilization in an ad hoc network with different environment settings and dynamic network changes (e.g., mobility and route changes).

Item Media

Item Citations and Data


For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

Usage Statistics