Doctor of Philosophy / Department of Electrical and Computer Engineering / Don Gruenbacher / Caterina Scoglio / In this dissertation, we investigate research gaps still existing in P2P file-sharing systems: the necessity of fairness maintenance during the content information publishing/retrieving process, and the stranger policies on P2P fairness. First, through a wide range of measurements in the KAD network, we present the impact of a poorly designed incentive fairness policy on the performance of looking up content information. The KAD network, designed to help peers publish and retrieve sharing information, adopts a distributed hash table (DHT) technology and combines itself into the aMule/eMule P2P file-sharing network. We develop a distributed measurement framework that employs multiple test nodes running on the PlanetLab testbed. During the measurements, the routing tables of around 20,000 peers are crawled and analyzed. More than 3,000,000 pieces of source location information from the publishing tables of multiple peers are retrieved and contacted. Based on these measurements, we show that the routing table is well maintained, while the maintenance policy for the source-location-information publishing table is not well designed. Both the current maintenance schedule for the publishing table and the poor incentive policy on publishing peers eventually result in the low availability of the publishing table, which accordingly cause low lookup performance of the KAD network. Moreover, we propose three possible solutions to address these issues: the self-maintenance scheme with short period renewal interval, the chunk-based publishing/retrieving scheme, and the fairness scheme. Second, using both numerical analyses and agent-based simulations, we evaluate the impact of different stranger policies on system performance and fairness. We explore that the extremely restricting stranger policy brings the best fairness at a cost of performance degradation. The varying tendency of performance and fairness under different stranger policies are not consistent. A trade-off exists between controlling free-riding and maintaining system performance. Thus, P2P designers are required to tackle strangers carefully according to their individual design goals. We also show that BitTorrent prefers to maintain fairness with an extremely restricting stranger policy, while aMule/eMule’s fully rewarding stranger policy promotes free-riders’ benefit.

P2P Grid computing seeks the convergence of Grid and P2P technologies. Deploying a P2P Grid middleware on a set of computers enables an organization to automatically barter computing time with other Internet-connected organizations. Such P2P exchanges of computing time enable individual Peers, i.e. organizations, to transparently aggregate large amounts of computational power with minimal infrastructure requirements or administrative cost. Challenges arise from the requirement for scalability and robustness. Individual worker nodes are unreliable, as P2P Grids operate on unmanaged desktop computers. A specificity of P2P Grids is that each Peer can reclaim at any time the computational power of worker nodes supplied to other Peers, leading to bursts of execution preemption. These are the major contributions of our dissertation: * Firstly, we propose a new P2P Grid architecture, the Lightweight Bartering Grid (LBG). Through systematic cooperation between Grid nodes, the reliability of execution of computational requests is greater than the sum of the reliabilities of worker nodes. * Secondly, we propose a highly scalable data transfer architecture. It is based both on the BitTorrent P2P file sharing protocol and on the removal of the temporal cost of downloading redundant copies of input data files. * Thirdly, besides a middleware implementation of LBG, we also provide an implementation of a discrete-event simulator. Its originality resides in the weaving of the simulator code into the bartering code of the middleware, which is made possible through the virtualization of Grid nodes. This enables reproducible testing and accurate performance evaluation of the bartering policies because the Peers of a simulated Grid make the same bartering decisions as Peers deployed on real computers. The LBG architecture exhibits the following remarkable features: * The scheduling model supports the queueing of external requests and the architecture enables a flexible study of bartering policies. * The architecture is open, flexible, lightweight and facilitates software engineering. It enables the easy development, testing, evaluation and deployment of combinations of scheduling policies. * The architecture is fully P2P.

Peer-to-Peer(P2P)架構讓社群內的使用者收集分散在網路各處之資源，其中最為風行的P2P系統當屬P2P檔案分享系統。P2P檔案分享系統之架構可分為集中式及分散式，而分散式架構又可細分為結構化及非結構化兩種。採用分散式且非結構化之BitTorrent-like架構，因其可擴張性較佳而廣為風行。在經驗中發現，BitTorrent-like架構在非對稱網路之下雖擁有寬裕的下行頻寬，但是其頻寬使用率卻不高，我們分析其成因如下：(1) Fractional Upward Bandwidth (FUB) 、 (2)Blockage of Acknowledgement (BoA) 與 (3)Long Physical Paths 等。而其中的Blockage of ACK問題，現今尚未有研究學者提出完整解決方案。本研究之目的，即要針對此Blockage of ACK的問題，改良網路協定中的傳輸層 (Transport Layer) 協定以提昇P2P檔案分享系統之效能。我們改用UDP作為傳輸層協定，並在應用層加入自動重建遺失之封包、決定基本傳輸單位大小及決定資料傳送速率等機制，以彌補UDP之缺陷。文中並提供了傳輸協定運作時所需參數的估計法，並且也與其它傳輸協定做效能之比較。實驗結果發現，我們設計的傳輸協定確可改善P2P檔案共享系統的運作效能。 / Peer-to-Peer (P2P) architectures let participants gather resources form network and make participants acquire more computation resources than they could offer. One of the most prominent P2P systems is P2P file sharing system. P2P file sharing system could be classified into to 2 categories: centralized and decentralized model. BitTorrent-like (BT-like) model can be classified as decentralized and unstructured model. BT-like model is quit popular nowadays due to its scalability. Unfortunately, BT-like model has several shortcomings on performance over asymmetric networks, because of some problems, such as Fractional Upward Bandwidth (FUB), Blockage of Acknowledgement (BoA) and long physical paths. No complete solution is available yet to solve the BoA problem. We propose a new UDP-based protocol to alleviate this problem. Since UDP protocol is not able to guarantee the data-integrity nor to determine the proper transmission rate by itself. To offer a complete solution, we have to enhance UDP, to add extra capabilities such as packet loss recovery, segment size determination and data rate determination mechanisms. Experiments show that our proposed protocol has good improvement on performance.