高韌性同儕網路及其在即時多媒體上之應用---在異質性同儕網路下即時多媒體內容之有效搜尋及繞徑技術

Abstract

由於VLSI/SoC 與無線通訊技術的快速發展，帶來低成本而具有高速運算能力的裝置及 大幅提升的有線/無線網路頻寬，使網際網路使用者之間的資訊分享越來越普遍。同儕網 路是一種在網際網路上的覆蓋型網路，它可以方便於資訊的分享，例如檔案與多媒體資 料(音樂與視訊等)的分享。在異質性同儕網路中，為了在大量的節點中有效率地搜尋到 想要的物件，專家已提出許多搜尋與查詢繞徑的方法。現有的研究方法中，通常會有搜 尋結果過量的問題，以及未將即時性與使用者相關內涵納入考量。在本子計畫中，我們 將設計並實作一種有效率的整合式垂直搜尋與查詢繞徑的方法，以適用於即時多媒體串 流資料的搜尋。藉由此方法，我們可以有效地在各節點中找到那些符合使用者查詢的物 件。除此以外，搜尋結果可包含每個目的地節點的特徵，例如平均流量與穩定度等資料， 以作為即時多媒體串流所需的參考資訊。我們會將位置與時間區域性納入設計考量，以 針對多媒體串流資料範圍，提出有效率的垂直搜尋方法。除此之外，我們還可以把搜尋 結果過濾並排序，以給予使用者最適當的搜尋結果。這部分可以利用即時多媒體串流資 料的領域知識與使用者相關內涵，如使用者之前的搜尋紀錄與使用者的喜好等。本子計 畫將會在三年內完成。在第一年中，我們將廣泛的進行相關文獻的探討，並提出一個能 將節點可利用率及即時多媒體串流的需求納入考量之有效的垂直搜尋與查詢繞徑演算 法，並模擬及評估其優缺點。在第二年中，我們將實作所提出的搜尋演算法並進一步改 良及評估其合用性。然後，我們將與其他五個子計畫做初步整合與測試，以作為搜尋演 算法之改進參考。在第三年中，我們將與其他子計畫做全面的整合與測試，並以查詢效 率與搜尋回應度來評估所提出之垂直搜尋與查詢繞徑演算法的整體效能。本子計畫的研 究成果可使同儕網路之搜尋更有效率及更精確地滿足使用者需求。隨著網際網路發展得 越來越龐大，以及消費者與企業越來越仰賴網際網路服務之際，本研究主題是一個很值 得深入探討的熱門主題。

With rapid progress of VLSI/SoC and wireless communications technologies, which brings low cost computing devices and much enlarged wired/wireless bandwidth etc., information sharing is more and more common among internet users. Peer-to-peer networks are overlay networks of the internet, which can conveniently benefit to information sharing, such as files and multimedia data (music and video, etc.) sharing. To efficiently search desirable objects among a huge number of peers in heterogeneous peer-to-peer networks, there have been many search and query routing approaches proposed. Existing approaches usually suffer from information overloading in search results and seldom take the real-time issue and the user』s context into consideration. In this subproject, we will design and implement an efficient integrated vertical search and query routing method that can nicely match with the real-time multimedia streaming data domain. By the proposed method, we can efficiently find those objects from peers that can meet the user』s query. In addition, the search response can include each target peer』s characteristics, such as average throughput, stability, etc., as reference information for better real-time multimedia streaming. We will design our approach based on an unstructured peer-to-peer network architecture with consideration of geographical/temporary locality. Also, we can discover, filter and rank search results, which will make use of real-time multimedia streaming data domain knowledge and the user context, such as previous search records and user preferences etc. to respond to the user with the most fitting search results. This subproject will be completed in three years. In the first year, we will extensively survey related search and query routing papers, propose an effective integrated vertical search and query routing algorithm with consideration of peers availability and the requirement of real-time multimedia streaming, and simulate and evaluate its virtues and weakness. In the second year, we will implement the proposed algorithm with further enhancement and evaluate its feasibility. Then, we will initially integrate with the other five subprojects for improvement and refinement. In the third year, we will fully integrate and test with the other subprojects, and evaluate the overall performance of the proposed vertical search and query routing method in terms of query efficiency and search responsiveness. The research results from this subproject can raise the efficiency and accuracy of peer to peer search in the vertical search domain, which is a hot research topic as the Web grows larger and larger, and customers and enterprises rely more and more on internet services.