| 摘要: | 隨著近年網際網路技術的快速發展與使用人數的大幅增加，傳統的網路架構已經無法應付如此龐大的使用量及變化量。為了降低龐大網路營運成本及更有效率的使用硬體資源，雲端運算 (Cloud Computing) 在此呼應而生。雲端運算是分散式系統 (Distributed System) 所延伸出的一種新概念，具有高擴充性、快速部屬、高可靠性及結合公用運算中的按使用量付費等優點。因為雲端運算擁有諸多上述優點，眾多的雲端服務提供者 (Cloud Service Provider) 如雨後春筍般出現，如Amazon EC2、Google App Engine和Microsoft Azure。 雲端運算目前乃在發展中，當使用者要存取不同雲端服務提供者的資源時，必須要向各雲端服務提供者的搜尋系統查詢，並沒有一個統一且有效率的搜尋機制。為了解決上述問題，本論文將提出兩種不同的機制，第一種機制，以分散式雜湊表為基底之多屬性範圍搜尋機制 (DHT-based Mechanism for Multi-attribute Range Query; DMMRQ)，讓使用者透過本機制能夠搜尋不同雲端中的資源。DMMRQ利用同儕網路 (Peer-to-Peer Network) 中的具有地域性的CAN (Content Addressable Network) 為基底來改良，具備分散性、高擴充性及防止單點故障等特性。除此之外，由網路模擬軟體PeerSim模擬結果顯示，DMMRQ的平均搜尋hop數較CAN改善31.43%，節點間的負載平衡改善24.71%。 而本論文的第二種機制為實作一個以代理人為基底之雲端搜尋機制 (Agent-based Cloud Query Mechanism) 於嵌入式開發板上，雲端代理人可替使用者來尋找不同雲端中資源的訊息。使用者只需要安裝一個應用程式且不需要去維護雲端資訊的清單和協定，即可輕易的索取不同雲端的資源訊息。在實驗中我們發現，當雲端代理人建置於家用閘道器嵌入式板上時，能夠同時承受大約300位使用者的連線數量，說明了雲端代理人的低負擔。 Due to the increasing amount of internet users and fast growing of internet technique in recent years, traditional internet structure cannot deal with such huge amount of users and change anymore. As a results, in order to reduce the huge cost of operating internet and more use hardware resources efficiently. Cloud computing comes into existence. Cloud computing is a new concept derived from distributed systems. It has contains advantages of high expansion, fast distribution, high reliability, and combining deciding payment by using amount of public computing. Thus, many cloud service providers has mushroomed in few years, such as Amazon EC2, Google App Engine, Microsoft Azure, etc. Cloud computing is still in development. When users need to access resources from different cloud service providers, user has to query the query system from cloud service providers; however, there is still no consistent and efficient query mechanism. Hence, in order to solve this problem, this thesis propose two different mechanisms, The first mechanism is a multi-attribute range query mechanism based on Distribution Hash Table (DHT) in cloud computing. By this mechanism, users can fast query resources from different clouds. The proposed query mechanism in this thesis enhances Content Addressable Network (CAN) with locality in peer-to-peer network, and therefore contains characteristics of distribution, high expansion, and preventing prevention of single peer breakdown. Furthermore, the simulation resulted by PeerSim show that the average query hop number of DMMRQ is faster than CAN by 31.43%, and the load balance between peers enhances 24.71%. The second proposed mechanism is Agent-based Cloud Query Mechanism (ACQM) implemented on embedded systems is the second proposed mechanism in this thesis. Cloud agent will search resource messages from different clouds in place of users. What users need to do is just install a application but doesn’t need to maintain the list and protocol of cloud information. Consequently, users can get resource messages from different clouds easily. By the experiment, show that when cloud agent is built on home gateway embedded board, it could sustain the online amount of about 300 clients at the same time. This represents the low overhand of ACQM. |
