Research On Cost-effective Monitoring Strategy For Multi-tenant Service-based Systems

With the rapid development of cloud computing and service-oriented paradigm,multi-tenant service-based systems(SBSs)have been widely used.A service system can provide services for hundreds of tenants,thus greatly saving service resources.In the cloud environment,SBS vendors compose web services(i.e,component services)in the form of business processes to meet the functional requirements and Quality of Service(QoS)requirements.SBSs are loosely coupled composed of multiple component services which are provided by independent(usually distributed)providers,various runtime anomalies may occur like failures of data transmissions.As a result,the system QoS may reduce.Service monitoring can be used to detect anomalies in a timely manner.However,with the large-scale application of multi-tenant SBSs in recent years,a large number of services need to be maintained,the monitoring cost rapidly increase.How to reduce the monitoring cost has become a huge challenge.Service monitoring consumes monitoring resources and incurs overhead,so it is unreasonable to monitor each service constantly and equally.If the same monitoring resources are provided for each service,it will not only cause a waste of resources,but also some services cannot obtain enough monitoring resources to ensure the stable operation of the system.In fact,the importance of each service is different.Therefore,in order to avoid unreasonable allocation of monitoring resources,different monitoring resources are allocated to services with different degrees of their importance,and it is also critical to develop a monitoring strategy for cost effectiveness by considering resource costs and system overhead while obtaining monitoring benefits.In addition,fulfilling personalized tenants' QoS requirements in monitoring is also significant.Each tenant's QoS requirement for their rented services may different.In the case of limited monitoring resources,all the services obtain the same monitoring resources,which will result in the unsatisfied QoS requirements for some tenants,and the violation of Service Level Agreement(SLA).In order to guarantee the QoS requirements of each tenant,this thesis ranks tenants and takes the rank as the key factor of service criticality,the service rented by the tenant with higher requirement is allocated with more monitoring resources to ensure the normal SLA.In SBSs,the time-aware service-based systems,such as the online video system,tenants have an urgent demand for response time.In order to reduce the monitoring cost and improve the tenants,satisfaction for their requirements of response time,we proposed the SLA-driven Cost-Effective Monitoring for multi-Tenant SBSs(SCEMTS).Due to the diversity of multi-tenant SBS,tenants in QoS-aware service-based system may have multiple-dimensional QoS requirements,in order to monitor the multi-tenant SBS with multi-dimensional QoS,We proposed Cost-effective Monitoring based on Multiple QoS Parameters of multi-Tenant SBSs(CMMSPTS).It can reduce monitoring costs for the multi-tenant SBS with multi-dimensional QoS,and improve the tenants' satisfaction for their personalized QoS requirements.The progress is completed from solving the problem of tenants'one-dimensional QoS requirements to the problem of multi-dimensional QoS requirements.The main works of this thesis are as follows:1.To monitor the time-aware response multi-tenant SBS,we proposed the SCEMTS.First,SLA parameters(including tenants' expected response time and the cost for expected response time in monitoring)are used to rank tenants.Then the service criticality is introduced to distinguish the importance of each service in the system.Service criticality is calculated by two factors,one is the response time-based criticality,second is the tenant-based criticality.The tenant rank is regarded as the main factor in tenant-based criticality,and the monitoring benefit based on the tenant ranks is defined.Finally,monitoring benefits,resource costs and system overhead are weighed to develop a cost-effective monitoring strategy to reduce monitoring costs and improve tenants'satisfaction for their response time requirements.In view of the lack of evaluation metrics for evaluating tenant satisfaction,the SLA violation rate is designed.And the degrees of tenant satisfactions for their QoS requirements under different monitoring strategies are compared by experiments.2.To monitor the multi-dimensional QoS multi-tenant SBS,we formulated the CMMSPTS.First,establishing a QoS-aware tenant ranking model,including tenants,multi-dimensional QoS requirements(such as response time,throughput,reliability)are classified and their values are standardized,the QoS requirements of tenants are comprehensively evaluated by linear weighting,so as to complete the tenant ranking.Then,Service criticality is calculated and the tenant rank is used as the main factor in service criticality.At last,monitoring benefit,resource cost and system overhead are weighed to develop cost-effective monitoring strategy,so as to reduce the monitoring cost and improve the tenants' satisfaction for their multi-dimensional QoS requirements.In addition,in order to simulate the multi-tenant SBSs of different business scenarios,comparative experiments of different QoS dimensions are set up,the monitoring strategy of multiple different tenants' QoS requirements are more comprehensively evaluated.In this thesis,the problem of cost-effectiveness and tenant personalized QoS requirements in the SBSs monitoring strategy formulation was researched in depth.We formulated two cost-effective monitornng strategies for multi-tenant SBSs.Experiments show that they are superior to other typical algorithms in meeting tenant SLA and controlling monitoring costs.It reduces the monitoring costs of SBSs in the cloud environments,and also improves the satisfaction degree of the tenants'personalized QoS requirements.It ensures the reasonable allocation of monitoring resources and maintains the stable operation of multi-tenant SBSs.