Research On The Key Problems Of Component Services In Cloud Service System

With the increasingly maturity of the cloud computing technology, the cloud service system has become an important model of software development. In cloud service system application, component services are deployed on different servers or server clusters. Because of the uncertainty of the user access behavior, when the user access traffic increases, the cloud service system can dynamically add component service replicas to improve the throughput of the system, and then ensure the performance of the cloud service system. Therefore, how to place and select appropriate component service replicas to guarantee the performance of cloud service system has become a current research hotspot.In this thesis, the relevant technology of component service replica in cloud service system is firstly analyzed, and three key issues are conducted including the number estimation, placement and selection of the component service replica. Aiming at the number estimation issue of component service replica, a component service replica number estimation algorithm based on throughput constraint is presented in this thesis. This algorithm guarantees the cloud service system throughput constraints according to the component service throughput aggregation rules which are established by the algorithm, and calculates the throughput constraint of each component service, then obtains the number of each component service replica. For the issue of the component service replica placement, a component service replica placement algorithm based on graph topology matching is proposed. In this algorithm, topological structures of the component services and the computing nodes are obtained by using clustering techniques, and the component service replicas are placed through matching the above two topological structures. This algorithm can effectively reduce the execution cycle and delay time of the cloud applications. In allusion to the issue of the component service replica selection, a component service replica selection algorithm based on grey-Markov forecasting is proposed through the combination of the load condition model and the grey-Markov forecasting technology. When selecting the component service replicas, this algorithm will forecast the load of computing nodes, and choose the computing node with the smallest load for component service replica selection. This algorithm can effectively improve the utilization rate of the component service replica. A simulation environment is built based on a cloud simulation software called CloudSim and a series of experiments are carried out. The simulation experiment results show the effectiveness of the proposed solution and algorithms.