Context-aware Offloading Strategy In Mobile Edge Computing

In the mobile edge computing environment,mobile devices offload some or all of their tasks to a suitable edge node for execution,limited by their own battery life and storage space.During the offloading process,contextual attributes such as user’s mobility,network state and server resources dynamically change with time,showing time-varying and fuzzy characteristics,so how to make efficient offloading decisions and provide low-latency,low-power and highly reliable services becomes a critical issue.The main work of the thesis is as follows.(1)To address the question of whether and where to offload tasks,we design a context-aware cloud-edge three-tier computing offloading architecture and a context-aware edge server selection framework.We use the multi-dimensional time-varying context in the mobile edge environment to make offloading decisions,and establish a time-varying model of the load,CPU utilization,network state,and user mobility of the candidate cloudlet(a kind of edge server).(2)To select the optimal cloudlet from multiple candidate cloudlets,we propose a cloudlet decision scheme of time-varying contextawareness(NSO)based on the neutrosophic set.The neutrosophic sets are developed from the fuzzy sets,which can better characterize inconsistent and imprecise time-varying information.We use the backward cloud generator algorithm of the cloud model to transform the time-varying data sequence of the context into a single-valued neutrosophic set representation,and perform a series of neutrosophic set operations to derive the best cloudlet.(3)User movement has a significant impact on the offloading decision.To measure user mobility comprehensively,from the periodicity of user movement,we propose a cloudlet usage prediction algorithm based on tail subsequence matching according to the historical access sequences of users to the cloudlet.In addition,the time-varying characteristics of user’s movement within the predicted cloudlet are comprehensively portrayed in multiple dimensions.Based on this,a mobility-aware cloudlet decision scheme(NSMO)based on the neutrosophic set is proposed.The experimental results show that NSO performs better than other context-aware offloading schemes and NSMO performs better than other mobility-aware offloading schemes in terms of number of failed tasks,response times and energy consumption.