一种多因素博弈的多租户资源动态分配策略

Multi-tenant Dynamic Resource Allocation Strategy Based on Multi-factor Game Model

多租户环境下工作负载的动态性要求云服务提供商能及时响应它们的资源需求,但是由于系统的可用资源是有限的,这需要有行之有效的资源竞争处理策略.针对这种需求,综合考虑租户SLA和租户的资源利用率,建立指导资源分配的博弈模型,通过动态感知和预测多租户运行时资源需求,结合租户SLA优先级和资源利用率进行在线竞拍,再利用博弈效用函数求解,得到最终资源分配策略.在此基础上,设计了租户资源预测、多租户资源竞拍、多租户资源分配等关键过程,实现基于动态负载的多租户应用在线资源分配,并通过实验验证了该策略的有效性.

In multi-tenant environments, dynamic workload requires cloud service providers to respond to their demands for resources promptly. As the available resources are limited, an effective resource adjustment strategy is needed. In response to this demand, based on the tenants＇ SLAs and resource utilization, it is necessary to establish a reasonable game model for resource allocation. By dynamically sensing and forecasting resource needs of multi-tenant at runtime, combined with tenants SLA priorities and resource uti- lization for online auction, a utility function is used to solve the game model, which can give a final resource allocation solution. The tenants＇ resource forecasting, multi-tenant resource auctions, multi-tenant resource allocation and other related key processes are de- signed, which achieves the resource allocation of online multi-tenant applications under dynamic workloads. The given experiments verify the effectiveness of the presented strategy.