Research On Optimization Of Dependent Task Scheduling Method On Edge Heterogeneous Platform

Edge computing,as an emerging computing paradigm,offers the advantage of migrating computational tasks to edge devices.Compared to cloud computing,it effectively reduces the latency introduced during data transmission,providing users with lower latency and a better experience.Currently,task scheduling is recognized as one of the key challenges in edge computing,and it has been a significant concern for researchers.However,most of the existing edge computing task scheduling approaches tend to treat all tasks within an application as a single entity or study them in a sequential manner,neglecting the interdependencies among individual tasks.Consequently,applications cannot be scheduled with finer granularity across different devices,which limits the flexibility and efficiency of task scheduling in edge computing systems.Therefore,in order to address the challenges of task scheduling in edge computing and meet the demand for low-latency applications in edge computing environments,this paper considers the dependency relationships between tasks and utilizes directed acyclic graph(DAG)to represent these relationships.Specifically,the paper analyzes the data constraints and interdependencies among tasks.The main focus of this paper is to investigate the problem of dependency-aware task scheduling in different scenarios.The primary contributions of this study are as follows:1)In the scenario of single-user and single-edge server,this research focuses on the problem of dependency-aware task scheduling in edge computing.The scheduling problem is modeled and analyzed,and its NP-hard nature is proven.A heuristic algorithm for dependency-aware task scheduling is proposed based on the priority relationships among tasks.This algorithm provides a task scheduling order based on the priority levels of tasks,ensuring the satisfaction of dependency constraints among tasks.It effectively addresses the problem of dependencyaware task scheduling by employing greedy and insertion strategies to optimize the scheduling objective and minimize task execution time.This approach enhances the efficiency and performance of task scheduling.Simulation experiments demonstrate that compared to other baseline algorithms,this algorithm significantly reduces the computation latency of applications.2)Considering the practical application environment where user devices may be covered by a heterogeneous computing platform consisting of multiple heterogeneous edge servers,users can select different types of edge servers for task offloading.In the case of multiple users and heterogeneous edge servers,this paper continues to investigate the optimization problem of allocating heterogeneous computing resources in edge servers and proposes a task scheduling decision algorithm based on an improved genetic algorithm.The shortcomings of the conventional genetic algorithm,such as limited exploration capability and premature convergence due to its simplistic crossover and mutation search methods,are addressed and improved in this study.The enhanced algorithm incorporates multiple search methods for generating new solutions during the crossover and mutation processes,effectively improving the algorithm’s global search capability and convergence speed.Finally,a series of simulation experiments are conducted to compare the proposed task scheduling strategy with other relevant works,validating the effectiveness of the proposed task scheduling algorithm in reducing task execution latency for user devices and improving task scheduling efficiency.