Multi-tenant Cloud-edge Workflow Scheduling With Priority And Deadline

Edge servers with limited computing resources are difficult to meet the needs of multitenant workflow application services,and often need to schedule some tasks to the cloud server.Different tenants have different priorities and deadlines.How to schedule tasks to maximize the overall quality of multi-tenant service is a key problem for cloud-edge service providers.The main challenges of this issue are as follows: 1)The overall service quality of multi-tenant is closely related to deadline and priority.The level of priority and the tightness of the deadline are contradictory factors that affect the quality of service.How to define the scheduling priority of tasks according to the deadline and priority is the key to maximize the Qo S of multi-tenant.2)Although the edge node is close to the user and easy to meet the task deadline,its resources are limited,and too many tasks will make some tasks overdue.Cloud computing is rich in resources,but it is far away from users,so it is easy to violate the deadline.How to allocate tasks with different deadline constraints to different resources to maximize the total service quality of multi-tenant is a challenge.3)A large number of active nodes of workflow instances have a large number of topological sequences.Multi-tenant multi-workflow instances greatly increase the number of topological sequences.How to find the optimal solution in the massive topological sequences is a difficult problem.For the problem under study,the priority-based multi-tenant instance success entropy is defined to measure the total quality of service of the multi-tenant.A level-based iterative greedy algorithm framework for workflow is proposed.The framework mainly includes scheduling sequence generation,resource allocation,initial solution generation,solution destruction and reconstruction,local search,acceptance criteria.In order to generate the optimal scheduling sequence,four tenant ranking strategies(the earliest tenant deadline first,the maximum tenant priority first,the minimum number of workflow instances first and the minimum workload priority first)and two task ranking strategies(the earliest average latest finish time first and the maximum rank value first)are proposed.In order to allocate resources reasonably,three resource allocation strategies are proposed,which are the earliest finish time first allocation strategy,the closest to the average latest finish time first allocation strategy and the resource allocation strategy based on tenant priority.In order to find the optimal solution,this paper proposes a level-based iterative greedy algorithm framework for workflow.Based on the different combinations of scheduling sequence generation strategy and resource allocation strategy,a variety of initial solution generation methods are obtained.A level-based destruction and reconstruction method is proposed,and a new scheduling sequence reconstruction method is constructed to guarantee the topological sequence.Based on the level-based neighborhood operation,the local search algorithm of variable neighborhood descent is constructed,and the feasible topological sequence is guaranteed.An acceptance criterion is proposed to accept the bad solution with a certain probability.To verify the performance of the proposed algorithm,this thesis uses multi-factor analysis of variance technology to calibrate the relevant parameters and components of the algorithm and selects the best parameters and components.The best algorithm is compared and analyzed with the other two classical algorithms.Experimental results show that the performance of the proposed algorithm is significantly better than that of the compared algorithm under different deadlines and different number of tenants.