由於使用社群服務(Social Service)的用戶數急遽上升，如何提供穩定的服務成為一個服務要點。對於社群服務而言，雲端環境可以提供靈活的運算能力及彈性的儲存資源。社群雲端資料中心必須儲存用戶的資料，供運算及互相存取，這些造成雲端資料中心巨大的網路流量，如何在存放相同資料的情形下改善內部網路瓶頸成為雲端資料中心網路優化的一個議題。 本論文主要包括利用社群平台Facebook提供的認證與授權介面實作出社群服務-Cyber Search Engine，提供以”人”為搜尋對象的服務，進而探討如何使用資料類別間的相依程度決定資料在雲端的放置位置; 利用k-means演算法提出配合雲端虛擬機之資料分群與放置方法，並透過模擬驗證其減少網路傳輸成本之效能。在社群服務實作方面，詳述使用到的技術及完整系統架構; 在資料放置方面，我們分別就虛擬機已固定在伺服器上，及虛擬機可任意放置於伺服主機上，分別提出資料放置方法。兩種方法均先將資料類型依相依關係及其被存取的次數轉為資料類型拓樸圖，拓樸圖中的點(node)代表資料的分類，線(link)代表資料分類間的相依關係程度，使用k-means演算法將此資料類型拓樸圖做資料分群以決定資料所適合放置的伺服器及虛擬機。模擬實驗結果顯示所提方法之網路傳送成本，皆比平均放置方法要來得節省。 ;It is challenge to provide a stable social service that can deal with a huge amount of information and users. Social cloud data center stores social related information generated by users and it causes processing bottleneck during operating the data flow of these data. Therefore, it becomes one of the critical issues to study the data placement issue so that the performance of the cloud data center can be optimized. In this thesis, a social service, which named “Cyber Serch Engine”, is developed by using Facebook login API and takes efforts to propose two k-means based data placement schemes to achieve better transmission performance in cloud environment. The social graph is adopted to represent the data dependencies and access frequencies. Thus, the link weight denotes the correlation degree between data types and the node weight represents the frequency of a data type being accessed.These proposes two data placement schemes, which are names as the pre-configured LXC scheme and dynamic LXC scheme, allocate social data in proper virtual machine and physical server depend on the relationship between data types to minimize the transmission cost. The architecture and technology of social service cyber search engine will be mentioned in detailed description. Then, simulations of the proposed two k-means based data placement schemes are provided. The simulation results show that both schemes illustrate better performance than the balance scheme.
