| Edge computing deploys edge servers at the edge of the network close to mobile devices.By offloading computation tasks from mobile devices to edge servers for execution,it provides mobile users with low-latency and high-bandwidth computing services.Workflow applications are widely used in mobile devices.The precedence relations among workflow tasks increase the complexity of task offloading.In addition,potential malicious attacks in the process of task offloading also pose challenges to the security of task offloading.Therefore,the design of efficient and reasonable task offloading algorithms for security-critical workflow applications becomes a key technical issue in edge computing.This thesis focuses on security-critical offloading of workflow tasks in edge computing.Considering the precedence relations among workflow tasks and the overhead of security services in the edge computing,a security-critical task offloading model with the optimization objective of minimizing the total completion time and energy consumption is formulated,in which a task sequence is regarded as a solution to the optimization model.To solve the formulated task scheduling problem,this thesis proposes a series of heuristics and metaheuristics to explore optimized task offloading solutions,in order to reduce the time and energy costs for task offloading as well as data encryption and decryption while satisfying the precedence constraints upon workflow tasks.Depending on whether the scheduling solution sequence satisfies the precedence relations among workflow tasks,task sequences can be classified into two categories,i.e.,precedence-aware and precedence-unaware scheduling solutions.With the satisfaction of precedence constraints,a precedence-aware heuristic and a precedence-unaware heuristic are designed to offload security-critical workflow tasks,respectively.Given a scheduling solution sequence and the operating conditions of edge servers,the heuristic algorithms select the currently best edge server to offload workflow tasks.Experimental results show that the precedence-unaware heuristic can achieve higher-quality scheduling solutions.On the other hand,due to higher searching capabilities of metaheuristic algorithms,this thesis further proposes a metaheuristic algorithm based on particle swarm optimization to solve the security-critical workflow task offloading problem.The proposed metaheuristic starts with an initial solution generation strategy to obtain a high-quality initial solution,and uses a mapping mechanism based on particle’s position information to convert each particle into a feasible scheduling solution.In addition,this algorithm incorporates the Levy flight strategy into the particle updating scheme and designs a new particle updating strategy with reduced computational complexity.Experimental results show that the metaheuristic can significantly improve the performance of security-critical workflow task offloading in terms of both solution quality and computational efficiency.Based on the abovementioned achievements,this thesis implements a visualized system for offloading security-critical workflow tasks.By integrating the security-critical offloading algorithms developed in this thesis,this system can simulate the procedure of offloading security-critical workflow tasks to edge servers and visualize the offloading results in terms of Gantt charts.The offloading algorithms and system implemented in this thesis are beneficial for improving the quality-of-service of edge computing as well as ensuring the security and reliability of task offloading. |
