Formulating Cost-Effective Tolerance Strategies For Multi-tenant Service-Based System

The popularity of the cloud environment has driven the development of a multi-tenant service-based system(Multi-tenant SBS),which simultaneously provides services for the tenants with multiple differentiated quality of services(QoS).This unique feature further complicates the problem of multi-tenant systems construction and adaptation.In a distributed and changeable cloud environment,software and hardware updates or server workload changes can cause dynamic changes in the quality of service of multi-tenant system components and lead to end-to-end quality violations.In general,a multi-tenant system reduces the impact of exceptions through applying runtime service adaptation(such as service reselection,resource reconfiguration,etc.),but this cannot adequately handle system failures.Meanwhile,choosing alternative services with equivalent functions to replace abnormal services may also lead to business interruptions in execution.Therefore,it has become a major challenge to select the right cloud services to build a multi-tenant system and quickly handle runtime exceptions.Considering the problems above,this thesis mainly makes contributions on the following aspects:First,it proposes a relationship evaluation algorithm for the conflict and cooperation between services in the process of cloud service provision,and constructs a multi-tenant system as a role-based collaboration(RBC)system.By analyzing the QoS quality,service cost,tenant service sharing and the collaborative relationship between services,the performance of component services is comprehensively evaluated,and then a multi-tenant system is constructed based on the performance evaluation of component services.The experimental results show that this method is more efficient than other representative methods.Second,in order to further solve the fault-tolerant ability of multi-tenant system in the case of exception,this thesis proposes an algorithm based on the dynamic service criticality to formulate a multi-tenant system's fault-tolerant strategy.As a precondition,system engineers of a multi-tenant system use the first method to select the appropriate component services tobuild a multi-tenant system.By analyzing the dynamic quality of component services,tenants' preference on multi-dimensional service quality and tenants' service sharing,we evaluated the service criticality and allocated redundant services for key services.The fault-tolerant strategy can replace the abnormal component services,so as to minimize the risk of system quality violations.