此篇論文提出一個以叢集為基礎的階層式同儕網路拓撲架構。此架構在龐大數量的參與節點中依然擁有可程式化之特性，且於高動態的網路環境中可自我調節。此拓撲架構的組成考慮了節點在網路中的實體位置關係與鄰近關係，且屬於同儕網路中的部份集中式與半結構化架構。在擁有多種特性的動態網路環境中，節點的不可預期性、移動性、搜尋效率及成本將成為此類研究必須克服之問題，然而本篇論文的設計可以讓此架構保有優良的效率與彈性。此外本系統中還加入了聲譽觀念以改善節點不願提出貢獻的問題。以聲譽為考量的特別階層式樹狀架構擁有將節點分群、分割及合併叢集，並減少不公平與不平衡的資源利用。另外，此階層式架構擁有一些緊密合作的程序，以確保在發生節點錯誤時擁有自我修復的彈性。至此，我們研究出了一個強健且有效率的階層式行動同儕網路拓撲架構。在實驗模擬中，此架構將在模擬程式中被驗證，且被證實具有在不降低網路效能的情況下，擁有對抗上述各種問題的能力。This paper proposes a hierarchical cluster-based peer-to-peer overlay scheme which is not only scalable to tremendous peer population, but also sustainable to high network dynamics in a large-scale network environment. This scheme is based on both partially-centralized and semi-structured overlay models by exploiting network locality and proximity. This joint design can keep up system efficacy and flexibility in dynamic network environments where several issues, peer churn, peer mobility, search redundancy and traffic overhead become stickier. In addition, reputation notion is integrated to mitigate the free-riding problem. According to special reputation tree hierarchy, the process of peer clustering elegantly operates cluster division and union to adjust the overall cluster hierarchy so as to somewhat moderate unfair or imbalanced resource utilization over the large-scale system. Furthermore, the hierarchical overlay is resilient to any single point of failure at any cluster. A tie-in process is able to link up the broken paths automatically. Therefore, our effort results in a robust, efficient hierarchical overlay in dynamic peer-to-peer networks. By simulation results, the proposed scheme can react to above issues without compromise of prominent performance.
