Research On Dynamic Resource Evaluation Model In Fog Computing

With the development of the Internet of Things and the increasing maturity of cloud computing technology,more and more devices are connected to the network,which makes the load of cloud data center increase,some delay sensitive services can not be timely response,and the Quality of Service(QoS)decrease.At present,most Internet of Things applications need to be responded quickly,some applications may involve data security issues.In this case,traditional cloud computing can not meet the requirements of these applications.Fog computing is an extension of cloud computing on the edge of the network,and has the advantages of low delay and high mobility.The data processing is close to the end user,which can perform local storage and computing.At present,fog computing has become one of the hotspots in the research of the Internet of Things.Due to the development of the Internet of Things,the needs of users for application services are becoming more and more complex,and the resources with the same service functions are gradually increasing.Resource management has become a challenge.How to meet the needs of users and ensure the utilization of resources is one of the main directions of fog computing.Therefore,the thesis chooses the resource estimation as the research topic and gives the method of resource estimation from the point of view of single user and multi-users.Considering the characteristics of user requirements and multi-QoS attributes of resources,the resource estimation method based on QoS and based on match game are proposed.The details are as follows.(1)A method of resource estimation based on QoS is proposed.The method considers the multi-QoS attributes and stochastic volatility characteristics of resources,and evaluates them by similarity matching and prediction method.Firstly,resources are classified and matched according to the weighted Euclidean distance similarity.Considering the constraints of resources,we introduce the penalty factor and the grey relational matrix to modify the similarity matching function.Then,we use the regression-Markov chain combination prediction method to analyze the load state change of candidate resources.In order to weaken the influence of the data fluctuation on the prediction results,the dynamic weight adjustment function of the sequence is added to optimize the data sequence.The experimental results show that the improved similarity matching method is superior to the traditional similarity method in the precision,and the regression-Markov chain prediction method can improve the prediction accuracy.(2)A method of resource estimation based on match game is proposed.The method takes into account the users and resource providers,and uses one to many two-side matching game method to model multi-users resource estimation and selection problem.In order to solve the conflict between resource attributes,we use VIKOR method to establish user requirements preference sequence for resources.Then we give the utility function according to the satisfaction of user requirements and resources to each other,use stable matching method to estimate all possible matching results and select the results with the maximum comprehensive utility function.The experimental results show that the method proposed in this thesis can guarantee the satisfaction of users and resource providers,and improve the overall resource utilization efficiency.