Research And Application Of Host Automatic Selection Strategy Based On The Cloud GIS Platform

Geographic information systems in all aspects of social life has been widely used, due to the high demand for the status of its processing power and spatial data difficult to buy, so that some SMEs and individuals do not have sufficient capacity to carry out development work related to geographic information. The emergence of cloud computing provides a reliable solution for these problems, so that users can purchase processing capabilities and spatial data on demand. Therefore, how to build a reliable cloud computing platform and provide users with efficient spatial data processing capability is the focus of the study.Virtualization is the foundation of cloud computing, and virtual machines as large-grained computing resources, migrating frequently among servers to achieve load balancing leads to unstable performance. Therefore, this paper uses load prediction method to predict server load conditions, and combines with a personalized resource allocation scheme to select the most suitable host to deploy a newly created virtual machine up, reaching a dynamic server load balancing.In this paper, the current mainstream prediction algorithms are analysed in a thorough way, and combining with geographic information system itself compute-intensive, memory-intensive features, this paper chose once exponential smoothing algorithm as the prediction algorithm and on this basis to improve the algorithm by extending the width of the original time-domain, adding a filtering process to eliminate instantaneous peak interference, enhancing the accuracy and stability of the algorithm. At the same time this paper proposes a personalized strategy to automatically select the host, the user will apply for virtual resources according to its own business demand and allocate each of the resource with a corresponding weight. Applications will be submitted to the management side, the management side will broadcast a signal to every node server, internal prediction algorithm gives prediction values, if the prediction value is higher than the threshold, this nod server response nothing, otherwise the node server sends fixed format information to the management side. The management side collects information and adds into the responseing list. Finally, in accordance with the weight, the management server traverses the responseing list to calculate the score for each node server, the server with the highest score is selected as the host server. Cloud GIS platform uses OpenStack open source project as the basis, combined with a wealth of functional modules of OpenStack to build a reliable cloud platform. Data storage uses the combination of Redis memory database and MySQL database. Geospatial data distribution system uses GeoServer. User interface uses the Django framework and the Python language to design a website, including a plurality of modules geographic information query, spatial data processing applications, virtual resources application and etc.This paper uses Selenium automatic testing tools and some monitor modules of Python to test the performance of load balancing. The results show that:the improved once exponential smoothing algorithm can show more real volatility and average server load condition, host automatic selection strategy in load balancing and stability of the system shows good performance.