| Abstract: | 愈來愈多人發現自己已經許久沒有翻閱紙本書籍，事實上，近年電子書及影音媒體以超過預期的速度成長，在可預見的未來，資訊web 化將對人們的生活產生革命性的影響。本計畫的目標是研發完整的動態內容應用程式的發佈、部署、更新等策略。工作項目分二年說明之：第一年階層式網路是最常見的網路服務遞送架構，本計畫設計實作階層式最佳化部署演算法(目前這是open problem)，並建置雲端內容服務。一個SOA 系統常出現群組與上下游資訊供應鏈的關係。傳統的單一程式部署問題並未考慮多重應用之間的整合。本計畫以服務部署矩陣表達多個內容服務者與多個服務提供者之間的關聯性，並研發新的演算法來求出最佳化多重服務部署策略。第二年在無時無刻變化的網際網路中，服務部署矩陣亦不斷變化。本計畫將一組服務元件的部署矩陣表示成單一服務元件的部署矩陣的組合(composition)，這能幫助部署矩陣的快速計算以及更新。利用雲端技術之MapReduce 機制，本計畫得以有效率地計算以及更新Internet-scale 服務部署矩陣。利用雲端技術，數位內容能夠有效率地從各個服務提供者取出並加以組合後遞送至使用者端。
Service-oriented architecture (SOA) is a computing paradigm uses services as basic building blocks to enable rapid and flexible development of software. This project aims to develop a Content Service System integrating the functionalities of digital document representation, content updating, and content delivering in the SOA system. Year 1 In first year, we build a content service framework by using the XML specifications such as WSDL, UDDI, and Service Choreography. In this project, a service application, comprising of many services, is represented by using the directed graph called SDG in which each vertex refers a service and each edge (a,b) refers to the dependency relation from service a to service b. In order to enhance the scalability and efficiency of the system, this project develops an efficient algorithm for deploying the nodes in SDG over the hierarchical network such that the total service operating cost is minimized. Year 2 A content service system usually contains of many applications with complementary functionalities. However, to our best knowledge, the integration of deployments of multiple applications has not been well-studied. In our project, the deployment matrix is utilized to represent the relation among multi-services and multi-hosts in the system. We develop the deployment matrix algorithm based on which the optimal multi-application deployment strategy could be found. Year 3 In a content delivery system, publish/subscribe is the fundamental operation to multicast digital documents among supply hosts and demand hosts. This sub-project develops a Router to support publish/subscribe operation in a Cloud computing environment. Relations to the Cloud computing are in two folds: (1) The massive deployment matrixes, developed in previous project, can be maintained via the MapReduce mechanism, and (2) Based on the deployment matrix, the collection of Routers cooperates and acts as a single high-performance virtualized data storage that fetch, combine, and deliver the contents between supply hosts and demand hosts. |
