The Research Of Resource Management Strategy Based On CGroup In Cloud Database

In order to ensure the quality of service, resource isolation technologies are becoming more and more popular in multi-tenant cloud database. Among them, CGroup can limit,record, isolate physical resources used by process group, which is lightweight, easy to deploy and management. Though isolation is implemented by CGroup, if resources can not be allocated reasonably, the system will generate a lot of resource fragments. In order to improve resource utilization, it is significant to study resource management strategy based on CGroup.Analyzing the weakness of existing resource management technologies and characteristic of cloud application, resource management strategy based on CGroup in cloud database is proposed, which considers the quality of service and resource utilization.Then the factors of SLA(Service Level Agreement, SLA) are defined, and based on these factors, database resources are allocated flexibly. After that, considering the amount of available resources and usefulness of remain resources, a strategy to allocate physical resources is designed. When dealing with tenant requests, whether to schedule or not is depending on the legality of requests. If schedule, the factors about CPU, memory, disk I/O and network are selected to evaluate workloads of server. When workloads are close,the tenant behavior pattern is extracted to evaluate the request correlation. And then,resource scheduling based on workloads and request correlation is implemented. Whether to consolidate or not is depending on servers' configuration and tenants' demand. If consolidate, in order to evaluate the complementarity of tenants' demand, the Demand Preference Timeline Vector is formulated based on the analysis of information recorded by CGroup. Then, a method of consolidation based on CGroup is given, which takes account of demand characteristic of tenants and resource configuration of servers.Finally, the effectiveness of resource management strategy based on CGroup isvalidated. The results show that the strategy based on CGroup can allocate resources flexibly to improve resource utilization, can schedule server based on multiple factors to improve efficiency, and also can consolidate resources based on demand complementarity to reduce competition when resources oversupply.