Energy And Latency Optimization On Dependent Task Offloading In Mobile Edge Computing

With the advent of the Internet of Things,massive data imposes a heavy burden on the bandwidth and throughput of the network.Meanwhile,there are real-time requirements for computing and communication,which is beyond the capability of traditional cloud computing.To address this issue,ETSI proposed the concept of Mobile Edge Computing(MEC for short)in 2014,which is a new platform that is close to users while providing IT services and cloud computing services within the wireless access networks.As the core technology of MEC,the performance of the computational offloading technology directly determines the quality of the service,and its important value has attracted the attention of the academic community.However,despite the endless emergence of relevant literature and emerging technologies in this field,little attention has been drawn to the dependent task offloading algorithm.In this thesis,we explore the dependent task offloading problem with deadline constraint in heterogeneous networks.We assume communication and computing resources are limited,which is common in real applications.Correspondingly,we propose an algorithm to solve this problem,which contains two phases: latency of application minimization phase(LM)and energy consumption optimization phase(ECO).First,in LM,we propose a rescheduling algorithm called Unfixed Start Time Dynamic Scheduling algorithm(USTDS),which innovatively transforms the assignment problem into the assignment-sort problem.Upon execution,USTDS determines the relative execution order between tasks rather than the exact start time of each task,thereby greatly increasing the flexibility of scheduling.Besides,the iterative attribute inherent in USTDS enables more room for adjustment in the task scheduling.Second,in ECO,we propose a local search optimization algorithm and a frequency modulation optimization algorithm,which convert the available delay to the corresponding reduction in energy consumption.In order to compress the solution space,we propose the "slack path" to avoid falling into the global search.We conduct a performance evaluation of the proposed algorithm under various DAGs.We evaluate the influence of the number of tasks and communication computation ratio in terms of communication delay and energy consumption.The results show that compared with the existing algorithms,our proposed algorithm is more efficient and suitable for the task offloading scenario in mobile edge computing.