Research On Multi-Tenant Resource Allocation Based On Game

In the cloud computing, the cloud service provider resources provide on-demand resource for cloud users through a network, such as computing, networking, storage, and other resources.In this way, cloud users’attention is no longer based on the cost along with specific computing device maintenance, but turned to services availability of cloud service providers. But the face of the huge pool of virtualized resources in multi-tenant, the question of how to manage and monitor the resource by tenants has become more prominent. At the same time, a tenant contains one or more virtual machines and has different resource consumption and service levels. And the average allocation method of resource management obviously cannot meet the resource needs of tenants. Also the tenant SLA violation may lead to the frequent occurrence, which can reduce the service satisfaction and improve the cloud service provider’s costs. Obviously this is not a situation that multi-tenants and cloud service providers would like to see. In order to manage the multi-tenant resource flexibly, the allocation of resources for multi-tenant has to consider the factor, such as the overall resource utilization of cloud service providers, the SLA of the tenant and the actual consumption of resources of the tenants.Firstly, to tackle the competitive problem of limited resources among tenants, this paper studies resource allocation strategy based on priority SLA tenants and tenants weigh bids and the auction game. This strategy takes advantage of kalman filtering to predict resource needs of tenants and sorts tenants according to their SLA priorities and tenant’s resource requirements for bidding strategies. Then, based on the game model and according to the game auction bidding algorithm, it can obtain the final tenant resource policy set. Finally, call resources adjustment interface for game successful tenants’ resources to allocate their resource. Experimental results show that, compared to the reward and punishment mechanism based resource allocation game, game model based on multi-tenant resource allocation strategies is better to meet the individual needs of tenants, and improve resource utilization of the system.Secondly, on the research of multi-tenant resource allocation strategies based on game model, multi-tenant resource monitoring and management of SLA prototype system is designed and implemented. The function of the system integrated tenants’SLA, tenants’application and tenants’resource monitoring and management. Finally, the system functionality and performance testing are implemented. And the results show that the proposed strategies are more adapted to dynamic load conditions and can better meet the individual needs of multi-tenant, to improving overall system resource utilization and protect the Qos demand of tenants.