The Study of Semantic-Based P2P Resource Sharing Models

由於網際網路的發達，使得資訊傳遞的速度越來越快，所能獲得的資訊也越來越多，該如何有效的查詢到所要的結果？傳統的做法是將資源集中在 Servers，其優點是能快速找到所要的資源，缺點是當使用者過多時會使得系統過載，造成整個系統癱瘓，因此就有學者提出使用P2P(Peer-to-Peer)來解決上述的問題。許多不同的應用都利用了P2P的概念，諸如檔案分享 (File Sharing) 、即時通訊 (Instant Message) 、以及分散式運算 (Distributed Computing) 等，而其中又以檔案分享最廣為應用。過去有許多的研究針對P2P系統提出有效的 application routing 機制來尋找資源，但是當分享的資源數量非常大時，這一類的研究無法保證一定可以找到真正所需要的資源。為了解決此問題，專家學者提出使用對資源做描述的方式，如此一來增加了查詢的彈性，也使得搜尋的正確性可以大幅提升，而且可以節省許多不必要的搜尋封包以及搜尋路徑而提升搜尋速度，但因為增加了對資源的描述，使得所需的儲存空間、傳遞的訊息量增加，造成網路頻寬的負擔。根據過去的研究成果，我們發現利用語意能夠有效的縮小搜尋的範圍，進一步增加搜尋的效率。在這個計畫中，我們希望能夠更進一步利用語意來提升搜尋效率，因此將針對兩個方向來做探討，一個是探討是否能更進一步利用語意間的關係來使得搜尋範圍變小、搜尋路徑變短。例如：一個利用廣播（Broadcast）方式傳遞訊息的P2P檔案分享系統，在查詢時是否能夠利用這些語意間的關係，將訊息往更有可能擁有資源的節點傳送，而不是盲目的傳送，使得搜尋的範圍縮小，增加搜尋的效率；另一個方向則是當這些資源描述被存在儲存庫（Repository）中，雖然能增加查詢效率，但維護成本也相對的增加。因此，如何降低維護成本及如何選擇適當的儲存庫將是我們所要探討的Due to the emerging of P2P technology, the concept of P2P has been widely used in many Internet applications such as file sharing, instance messaging, and distributed computing. As a result, many researches had been devoted to improve the efficiency of P2P application routing protocols. Protocols such as Freenet, Pastry, and Chord had been proposed. However, this type of research cannot ensure the queried resources can be found when the amount of shared resources is huge and their diversity is high. To resolve the problems, semantic-based P2P systems were heavily studied in recent years. In semantic-based P2P systems, each shared resource is described by some metadata. With metadata, the resources can be correctly located. In addition, because application routing protocols can take advantages of these metadata, the number of query messages can be reduced. Also, only the “possible” routes will be used. Although many claimed the usage of metadata will significantly increase the overhead, we believe that the overall performance of semantic-based P2P systems can be further improved based on our previous research results. Therefore, in this project, we aim at two approaches to further take advantages of semantics. The first approach is to investigate whether or not the semantics can be further utilized to reduce search space either in broadcast or DHT manner. For example: A P2P file share system which transmitted the message using broadcast, we can not only use relationships between the resources but also derive or inference relationships from existing relationships to restrict flooding space. The other approach is that, when these metadata are stored in repositories, how to select appropriate repositories to increase query efficiency while maintenance cost can be reasonably low