可分割云资源下多租户效用最优化策略

摘要

For efficiently solving the competition allocation problem of divisible cloud resource among multi-tenants,an optimal allocation strategy based on utility function was presented.The evaluation function and cost function of resource allocation were defined and utility optimization model was built.The effects of different cost functions on optimal allocation were focused,and the fact that a unique Nash equilibrium solution of resource allocation existed in this utility model was verified.The performances of resource allocation in different forms of cost functions were compared by simulation experiments.Experimental results show that linear and convex cost function can converge to unique solution and linear cost function can bring optimal system efficiency for single tenant and maximize the overall social welfare of multi-tenant in the process of social welfare maximization.%为有效解决多租户对可分割云资源分配的竞争问题,提出一种基于效用函数的最优化分配策略.通过定义竞争租户对资源分配的评估函数和代价函数,建立效用最优化模型.讨论不同形式的代价函数对最优化分配结果的影响,验证在该效用模型下资源分配存在纳什均衡解,仿真对比不同形式代价函数下资源分配的性能表现.实验结果表明,在求解租户社会福利最大化过程中,线性代价函数和凸性代价函数均可以使分配过程收敛到唯一解上,但线性代价函数可以为单个租户带来最优的系统效率,实现多租户环境下整体社会福利的最大化.