| Summary: | 碩士 === 輔仁大學 === 資訊工程學系 === 99 === In cloud computing, there is a huge amount of data in storage devices. It is an important issue how to ensure the data continues to be available when failures occur in storage devices. Data replication is a widely used technique. This thesis presents an efficient data replication algorithm for cloud computing systems. The algorithm considers the QoS requirements of applications (e.g. response time) and the heterogeneous characteristics of devices (e.g. different transmission rates, data access time, storage capacities, etc). There are two stages in the proposed algorithm. In the pre-processing stage, data replication requests are modeled as a weighted bipartite graph for considering QoS requirements and device heterogeneities. In the post-processing stage, the weighted bipartite graph is extended as a flow graph. Using the flow graph, the data replication problem can be optimally solved by transforming this problem to the well-known minimum cost flow problem. Finally, this thesis performs simulation experiments to demonstrate the effectiveness of the proposed algorithm in the replication cost and recovery time.
|
|---|
