Scheduling algorithm research for spatial information service under cloud environment

It is well known that the application of Spatial Information Service has a problem of computation intensive. In traditional solution, it is almost impossible to estimate the computing resources of the application and thus either waste or shortage of resources will appear at last. Aimed at it, this paper combines the cloud computing features and presents an algorithm called PSA (Prediction Scheduling Algorithm) to balance the resources and performance. The approximately infinite scalability and rapid deployment ability of cloud computing depicts the foundations of PSA, while the analysis of user request gives the guidance to PSA. Combines with above two keys, the PSA can dynamically adjust the resources used in the application with the prediction from guidance. Later on in this paper, an elaborate experiment under strict control in real Computing Environment is held with four different types of user requests (uniform, unimodal, symmetric bimodal and asymmetric bimodal) to simulate the performace of PSA under distinct situations. For comparison reasons, there are another two algorithms rasing called SCA (Single Scheduling Algorithm) and TISA (Threshold Incremental Scheduling Algorithm), which represents the traditional architecture of Spatial Information Service and simple utilization of cloud computing respectively. The experiment data shows that PSA can largely improve the system performance and return of investment in terms of cost and user experience under concentrated user request type and slightly improve in uniform situation. Therefore in current research stage, PSA is the best algorithm to balance resources and performance.