Research On Micro Cloud Structure Optimization And Computing Offload Based On User Interest Model

Since mobile devices are widely used in recent years,mobile cloud computing technology develops rapidly as well.Nowadays,mobile devices can access cloud resources like fixed devices through wireless network.The routers receive mobile user requests and transfer them to cloud.However,as mobile devices are irregular and dynamic comparing with personal computers,which will result in higher network delay and larger transmitting energy consumption.In particular,when network is extremely unstable,it will lead to problem like increasing probability of failed requests.Due to the large scale and centralized location,cloud data centers also face problems such as poor scalability,low resource utilization,high cost on equipment and management,and paid services.With the development of data-intensive applications,mobile users require less response time and higher service quality.Thus,the execution efficiency of mobile data flow application should be improved to meet user demand.As applications consume huge amount of power and mobile devices have limited battery capacity,these would also decrease user experience.In view of the above-mentioned problems,the main contributions of this article can be described from three aspects.(1)In order to reduce network delay and take mobile device features into account,the micro-cloud architecture based on user interest is proposed.We analyze footprints of mobile users when they surf online or use applications.According to the mobile user interest and mobile device location,the node interactive model is established to make maximum use of cloud resources and shorten response time.The architecture is related to completion time and service satisfaction of each node.These two parameters are used for calculating measurement level of each mobile device.The requirements of node connection are also discussed in the article.With node measurement level and connection requirement,the micro-cloud architecture can be established.(2)Since mobile devices consume huge amount of energy,that make it difficult to process numerous multi-tasks for a long time.Additionally,mobile users require high quality of user experience,which means the response time should be reduced.Therefore,the offloading strategy for proposed micro-cloud architecture is presented.The data flow application model is built up by analyzing offloading process of mobile device,application features and relationship between task quantity and application components.Moreover,the mobile offloading performance model includes wireless communication model,response time model of data flow application and mobile device energy consumption model.The offloading strategy based on bacterial foraging optimization algorithm is proposed for reducing energy consumption and response time.(3)The experimental verification has two parts: micro-cloud architecture optimization experiment verification and offloading experiment verification.In micro-cloud architecture optimization experiment,parameters of mobile user interest model are verified,such as semantic relevancy threshold,damping coefficient and feature dimension.Then the improved feature extraction algorithm is compared with three other typical feature extraction algorithms,and interest feature classification algorithm is compared with Bayes algorithm,SVM algorithm and KNN algorithm.The simulation results show that the proposed algorithm has optimal values on macro average accuracy,macro average recall and macro average F1.Meanwhile,the validity of proposed micro-cloud architecture is demonstrated from two aspects: stability and scalability.The offloading experiment verification uses variable control method.The proposed offloading strategy is compared with other existing offloading strategies under different conditions.The response time and energy consumption are compared respectively under different mobile device conditions,different task quantity conditions and different sizes of micro-cloud architectures.The simulation results show that the proposed offloading strategy can reduce more response time and save more energy than other algorithms.