Doctor of Philosophy
School of Electrical, Computer and Telecommunications Engineering
Raheel, Muhammad Salman, Resource management In P2P and MANETs using layered coding techniques, Doctor of Philosophy thesis, School of Electrical, Computer and Telecommunications Engineering, University of Wollongong, 2015. http://ro.uow.edu.au/theses/4590
The thesis addresses the resource utilization problem to provide smooth video delivery over P2P wired or Mobile Ad hoc networks by exploiting the properties of layer coding techniques. Given a request for a video, the aim is to stream the video to all destinations with the maximum achievable quality. This problem is known as the resource utilization problem and it has been studied extensively over wired and wireless networks, and it is known to be NP hard. However, due to the high dynamicity of peers and current network conditions, the resource allocation problem is still open. Most of the carried research considers streaming the videos to the destination node using a single source and without implementing any coding techniques which introduces huge playback issue. Moreover, in the context of MANETs, the resource utilization adds further challenges as nodes are considered to have limited energy with a highly dynamic topology. Recently, there has been much research carried out towards providing different routing protocols or streaming techniques to efficiently handle the resource. However, most of the work considers either energy or link bandwidths as a constraint to handle the resources of nodes.
This thesis, investigates the resource utilization problem over P2P and MANETs using layer coding to efficiently utilize the available resources in the network. Hence, the thesis proposes approximation algorithms to this purpose. The main contributions of the thesis are summarized below;
The thesis proposes an algorithm that exploits the properties of Scalable Video Coding (SVC) in order to minimize the upload bandwidth at each peer. More specifically, the concept of streaming different layers of the same video from different peers has been proposed. Further, an optimization problem is defined to handle the upload bandwidths at peers. However, the solution to the proposed problem is NP complete. Therefore, an approximation algorithm is proposed to solve this problem. In addition, seed servers are introduced in order to deal with extra load in the network. The proposed method provides better performance as compared to the current approaches that use single layer video in combination with SVC. The simulation results are compared against the model proposed in the literature. According to the results, the proposed model improves diversity, increases average video quality, reduces the effect of churn and manages flash crowds.
Apart from basic P2P network, the thesis also investigates the resource allocation problem for distributing the video in a P2P mobile ad hoc network (MANET) to provide users' with a better quality of experience (QOE). Therefore, a linear optimization problem to efficiently utilize the upload bandwidth at each mobile node is defined. Scalable video coding (SVC) is used to help maximize the Quality of experience (QOE) by distributing the load across the nodes to minimize the power consumed and the upload bandwidth at each peer. However, the solution to the proposed problem is NP complete. Therefore, a QOE based Energy Efficient model (QEE) is proposed that provides an approximation algorithm and compares the performance of the proposed model with the existing models as explained in the literature. The simulations results show that QEE provides better QOE, consumes less power and minimize the upload capacity at each peer as compared to the existing models. Furthermore, QEE model also helps to manage the flash crowd and effect of churn in the network.
The thesis also addresses the data collection and routing problem for streaming video over a decentralized MANET to improve the average video quality received. The solution to such a problem is known to be NP complete. Hence, a novel Energy- Efficient Video Streaming method, called EEVS, is proposed that provides an adaptive data collection technique and a routing protocol to share the video across the network. In adaptive data collection technique, the nodes share their available information across every node they meet. However, the routing protocol helps to identify the sources and stream the video through multiple sources towards a given destination to reduce the overall load at each node. Furthermore, in order to handle the heterogeneous peers in MANETs, Multiple Descriptive Coding (MDC) is used which provides the video at different quality levels. The performance of EEVS is compared other well-known protocols in two experiments. In the first experiment, the data collection phase of EEVS is compared against MVSS and HAS-A-GEM. In this experiment, the information available across the nodes is shared across every other node they meet. In second experiment, the routing phase of EEVS is compared against EDSR and MP2P+MDC. The simulation results show that EEVS has 120% less overhead than HAS-A-GEM and approximately 170% less overhead than MVSS. Furthermore, the results show that the EEVS outperforms MP2P+MDC and EDSR by efficiently managing the energy across the nodes and distributing the load across the network using multiple sources. Hence, this increases the network lifetime. Moreover, the results also show that in EEVS the average video quality received is 30% more than MP2P+MDC and approximately 50% more than EDSR. The results also show that EEVS reduces the streaming delay up to 165% as compared to MP2P+MDC and approximately up to 300% as compared to EDSR.
This thesis is unavailable until Tuesday, July 31, 2018