Research On DAG Scheduling With Deadline-constrained Strategy In Cloud Computing

With the increase of large-scale data computing problems in many fields,cloud computing based on distributed computing model has attracted the attention of scientific research institutions at home and abroad.It uses virtualization technology to integrate the resources of data center and provides computing,storage and other services to users through the network.Users pay a certain fee to obtain resources according to their own needs.Cloud computing,as a new business computing mode,is more concerned about the quality of service provided to users in the specific execution process of user tasks.Reasonable task scheduling strategy can improve user experience in all aspects.This paper elaborates the related task scheduling technologies in cloud computing field,classifies them according to different processing methods,and summarizes the advantages and disadvantages of various scheduling algorithms.At the same time,aiming at the shortcomings of current DAG task sub-deadline acquisition technology and cost optimization in multi-DAG environment,two improved algorithms are proposed,which can effectively guarantee the completion of users' tasks within the deadline and improve the fairness of users' task execution costs.The main innovations of this paper are as follows:(1)To overcome the shortcomings of existing scheduled deadline acquisition methods for Deadline-Constrained DAG tasks,the RHEFT algorithm is improved.The scheduling sequence of DAG tasks based on HEFT algorithm is inverted,and the prescheduled tasks are inverted from the back to the front,and the sub-deadlines of tasks are calculated by the pre-scheduled results.Experiments show that the improved algorithm makes the task sub-deadline acquisition more accurate in DAG.(2)In view of the fairness of the existing multi-DAG task scheduling cost optimization algorithm with time constraints,the MRHEFT algorithm and CORM algorithm are improved,and the relative cost of each user is measured by combining the single DAG cost optimization scheduling algorithm with the idea of DAG emergency degree.Users with longer deadlines will get more preferences.For users with too urgent deadlines,task execution costs will rise relatively,and there is a risk that DAG tasks will be discarded,which meets the expectations of cloud service providers that users can set relaxed deadlines.The algorithm first obtains the subdeadlines of tasks in multi-DAG environment by improved MRHEFT algorithm,then takes the execution cost of DAG running in the system as the constraint point,and makes the running cost of all users' DAG at an average level by retrospective reset method,which improves the fairness of the cost among users.