在近年來P2P(peer-to-peer)系統逐漸被用於網路上分享各種資源,尤其是應用於資料量較為龐大的資訊分享。而在龐大的資料量中,如何要有效率地找出所需要的資訊變成一個重要的議題。在於分散式非結構的P2P網路架構中,Gnutella,是一個最基本的搜尋系統,主要是使用flooding來進行搜尋,但是這個方法會造成網路上流量太多不必要的消耗,因此後續又演發出random-walk、cache等方法,來減少流量消耗。 但是以上這幾種方法都在於流量消耗以及搜尋效率這兩種主題打轉,而忽略了這種搜尋系統主要還是由使用者來發出查詢,因此基於小世界網路模型(small-world network model)的概念特性下,本篇論文提出了依使用者查詢時的興趣特性,來調整其整體網路連結,試圖來縮短搜尋路徑以及時間,並且減少網路的流量。
In recent years, P2P (Peer-to-Peer) systems are widely used to share and exchange files, especially for large ones, between peers. How to efficiently locate peers that are relevant for a given query is a central issue in P2P systems. In unstructured P2P networks, Gnutella is the most famous P2P system. Gnutella would waste a lot of bandwidth, since it uses flooding to locate and search resource. Furthermore, random-walk and cache were presented to reduce bandwidth and improve search efficiency. But all of these systems ignore the user’s background such as user’s interest and common behavior. Based on the small-world network model, this thesis proposes a new approach, which allows peers to dynamically change the topology according to user’s interest. Experimental results show that the proposed approach outperforms the prior works in terms of hit ratio, routing path, or packet flows.