Research On Efficient Computational Offloading Oriented To Dependent Tasks

With the rapid development of the Internet of Things and artificial intelligence technology,various terminal devices emerge at the historic moment,resulting in the explosive growth of user data.New applications are increasingly computation-intensive and delay-sensitive,posing serious challenges for resource-constrained end-devices.As a new computing paradigm,mobile edge computing(MEC)provides users with offloading and computing services by sinking computing and storage resources in the cloud center to the edge of the networks,which fundamentally reduces the delay of data communication and has been viewed as a promising technology.At present,the research on mobile edge computing networks mainly focuses on independent task scheduling,but the research on dependent tasks is very limited.For example,the execution of one subtask depends on the output of another subtask,and some subtasks can be offloaded to the server for execution while others can only be executed locally.Firstly,task dependencies and the constraints of hardware and software will have a great impact on task scheduling and offloading.Secondly,the existing studies on task scheduling and offloading focus on single task response delay or energy consumption without considering them jointly.Because of the above phenomenon,this thesis firstly models the dependent tasks and proposes a new task scheduling and offloading scheme based on this model,which jointly optimizes the energy consumption and the delay.Then the scheme is improved by jointly considering the task response delay and device-end energy consumption.The main content of this thesis is as follows:1.This thesis firstly studies the task structure of existing Internet of Things device applications,and the study shows a complex logical relationship between the various parts of many tasks.In this thesis,the task is modelled more realistically based on the Direct Acyclic Graph(DAG)to provide a basis for scheduling offload research that considers the internal dependencies and hardware and software constraints.Considering a more realistic network,this thesis studies an ultra-dense edge network system in which terminal equipment in a user service area can communicate with multiple heterogeneous servers.In this system,based on the above task structure,a task scheduling and offloading scheme are proposed,which considers both the task response delay and the energy consumption of the terminal equipment.The scheme divides the task scheduling offloading process into two steps: priority establishment among subtasks and subtask offloading.The priority establishment among subtasks is to obtain the subtask priority sequence through the recursive operation on the task cost statistics.In subtask offloading,redundant computing is introduced,and the task response delay and energy consumption are jointly optimized to decide each subtask to the corresponding server.The simulation results show that the proposed scheme has a lower task response delay than the existing computing offloading schemes from the perspective of performance,and the task model and system model have more realistic physical meaning from the perspective of the system.2.Aiming at the above-mentioned ultra-dense edge networks,dynamic voltage frequency adjustment is introduced in this thesis based on the joint consideration of task response delay and device-side energy consumption to improve the proposed redundant computing task scheduling scheme.By rationally selecting the device-side CPU task execution frequency,the total energy consumption of the terminal device is reduced on the premise of keeping the task response delay unchanged.Simulation delay results show that the introduction of dynamic voltage frequency adjustment can reduce the energy consumption of equipment and increase endurance while ensuring the performance of task response delay.