Multi-tenant Instance-intensive Workflows Scheduling With Resource Reliability

Multi-tenancy is one of key features of cloud computing,which provides resource sharing for economies of scale.The rapid development of e-commerce makes instance-intensive workflows scheduling problem in a multi-tenant cloud increasingly prominent,and it also leads to the problem of resource reliability in high-concurrency environments.Resource failure will result in failure of task execution and affect scheduling performance.In this thesis,it is of great theoretical and practical value to study multi-tenant instance-intensive workflows scheduling method with resource reliability.In this thesis,the multi-tenant instance-intensive workflows scheduling problem with resource reliability is studied.A modified workflow scheduling reference architecture in a multitenant cloud environment is proposed,and resource reliability service module is integrated into it,which enables multiple tenants to run workflow instances on a shared cloud infrastructure.In view of the important impact of resource reliability on workflow scheduling problem,a hybrid resource reliability assessment method is proposed.This method is used to evaluate resource reliability in real time.According to the characteristics and nature of the problem,the concept of instance success entropy is applied to evaluate the QoS of all tenants.The smaller the instance success entropy is,the better the QoS of the application is.A mathematical model is established whose objective is to minimize the instance success entropy of tenants.Based on the above,a workflow scheduling with resource reliability algorithm(WSRRA)is proposed.The algorithm mainly includes four important parts: QoS division method,sorting rules,task scheduling and resource reliability updating method based on time window.The QoS division method consists of a sub-deadline assignment method and a sub-budget assignment method.Sorting rules mainly include tenant prioritization rules,task sorting rules,and resource sorting rules.Ready tasks are scheduled in a time window.Resource reliability is an important consideration when selecting resources to improve tasks and instances execution success rate.During the task scheduling process,the resource reliability updating method based on time window is used to update the resource reliability in real time.To verify the effectiveness of the proposed algorithm,the analysis of variance(ANOVA)technique is adopted to carry out parameter calibration and component analysis and the best parameters and components are selected.The algorithm proposed in this thesis is compared with the related algorithm of this problem.The experimental results show that the proposed algorithm is obviously superior to the comparison algorithm under the deadline and budget constraints of different level.