Research On Container-Based Edge Computing Task Scheduling Method

As the internet of things industry rapidly develops and user standards for services gradually improve,the shortcomings of the traditional cloud computing model of transmitting all tasks to the cloud center for processing have become increasingly prominent.Edge computing,which is an extension of cloud computing,has garnered considerable attention due to its potential to efficiently reduce network congestion and transmission latency,and enhance service quality.However,edge computing devices are limited by factors such as environment or energy consumption,and their device resources are generally limited.Thus,how to better utilize these limited resources to provide services has become a challenging problem.In addition,current research generally categorizes all tasks into one category for indiscriminate processing.However,different tasks require different services,and only after configuring the appropriate services can these tasks be processed.Therefore,this article focuses on the research of task scheduling under the container service configuration model,and the main work is as follows:1)The task scheduling model with container constraints has been established.To address the difference in task service requirements while allowing limited resources to process more types of tasks,this article introduces a container service configuration model in edge nodes,establishes a matching relationship between tasks and container services,and conducts research on multi-task independent scheduling scenarios and multi-task dependent scheduling scenarios.The corresponding task scheduling models are established,and task scheduling algorithms are proposed.2)The task scheduling algorithm that improves the independence using a multi-strategy improved genetic algorithm has been proposed.For the independent task scheduling scenario,this article first uses task execution latency and system energy consumption as evaluation objectives,and constructs time,energy cost,and task container matching constraints to establish a container-constrained independent task scheduling model.We propose an enhanced genetic algorithm to address task scheduling,which leverages multiple strategies to optimize and improve upon traditional approaches.After the mutation stage,individual optimization is performed using the Discrete Beetle Algorithm,which randomly selects and replaces the scheduling scheme in the population.The experimental results indicate that the algorithm achieves a substantial reduction in the optimization objective while ensuring rapid convergence.3)The dependency task scheduling algorithm based on the Double Dueling Deep Q-Networks(D3QN)has been proposed.To address the scheduling of dependent tasks in scenarios involving multiple tasks,considering the execution order and impact of dependent tasks on the optimization objective,this paper proposes a topological sorting algorithm.The algorithm prioritizes the processing of smaller tasks to ensure critical tasks are handled first,thereby reducing the wait time for subsequent dependent tasks during the task scheduling process.Additionally,we use the D3 QN to handle dependent task scheduling,treating it as an agentaccumulated reward process,and design a suitable reward function and system state to optimize action selection through accumulated experience,thereby providing an appropriate scheduling plan.The simulation results demonstrate the favorable performance of the proposed algorithm in effectively managing dependency tasks.