| In recent years,the rapid development of information technology has led to a sharp increase in the types and number of embedded devices,which puts forward higher requirements for the development and testing technology of embedded software.Traditional testing methods are not only limited by region and time,but also have the problems of complex deployment of testing environment and high cost.The combination of traditional embedded system testing and cloud computing technology can integrate decentralized testing resources and provide users with a remote access,convenient and efficient new testing service mode.The performance of task scheduling strategy in embedded system cloud testing environment directly affects the efficiency of test cloud,resource utilization and user experience.The existing task scheduling methods in cloud environment lack of multi factor consideration of inter task dependency,scheduling efficiency and user service quality.Based on this,this thesis focuses on the task scheduling method of embedded system cloud testing.The main contributions include:(1)Aiming at the scenario of associated task scheduling in embedded system cloud testing environment,this thesis proposes an improved genetic algorithm(UEGA)based on user expectation.Firstly,the directed acyclic graph is used to model the dependencies between tasks,and then the heterogeneous earliest completion time algorithm is used to deal with the dependencies in the initialization stage of genetic algorithm to form a multi priority queue.Finally,the crossover and mutation operators suitable for the scheduling scenario of directed acyclic graph are constructed,at the same time,the dynamic pricing model is introduced into the fitness function to dynamically adjust the actual cost according to the waiting time of users,so as to improve the service quality of users.Comparative experiments on Cloud Sim cloud simulation platform show that UEGA shortens the completion time of tasks and improves user satisfaction.(2)Aiming at the problem that the genetic algorithm has local convergence and can not effectively use the feedback information in the later stage,this thesis further combines the UEGA with the ant colony algorithm with positive feedback mechanism,and uses the better solution obtained by the UEGA to initialize the pheromone of the ant colony algorithm,so as to solve the problem of low solution efficiency caused by the lack of pheromone in the initial stage of the ant colony algorithm.The load balancing factor is introduced into the improved genetic ant colony scheduling algorithm,and the calculation method of transfer probability matrix and pheromone update rules in ant colony algorithm are optimized to improve the performance and load balancing degree of the algorithm.Compared with UEGA and ant colony algorithm,the improved algorithm effectively reduces the task completion time and cost,and improves the overall load balance.(3)The task management module of embedded system cloud testing platform is designed and implemented.Firstly,based on the overall architecture of embedded system cloud testing platform,the functional requirements and R & D technical route of task management module are determined.Then,the task management module is briefly designed and implemented in detail around the module architecture,database and important function implementation process,and the improved scheduling strategy is applied to the task scheduling sub module to improve the scheduling efficiency of the cloud platform.Finally,the function of the task management module is tested to meet the functional requirements of the task management module. |
