移动云中基于随机博弈的多用户任务卸载效用优化

Multi-user task offloading utility optimization based on stochastic game in mobile cloud

针对移动云中多个资源需求节点(resource demander,RD)同时卸载任务到资源提供节点(resource provider,RP)时会造成各个RP上任务分布不均匀,从而导致节点任务执行时间不一致引起移动云服务效率下降等问题,提出了基于随机博弈的多用户任务卸载方法。该方法以任务卸载节省的时间为各个RD效用,建立基于节点效用最大的优化模型;分析该模型并依据RD中任务产生的随机性,将节点效用优化问题转化为每个RD为获取最大效用而进行博弈的问题,确定该博弈为N玩家随机博弈,提出能够达到ε-纳什均衡解的反向迭代算法以得到每个RD的纳什均衡策略;并将获得的策略与采用静态博弈得到的策略进行实验对比,结果表明,采用随机博弈得到的卸载策略能够有效降低任务执行时间,同时提高移动云服务效率。

When multiple resource demand nodes(RDs)offload tasks to resource provider nodes(RPs)simultaneously,such situations as inhomogeneous tasks distribution,unstable execution time and the low service efficiency of mobile cloud can be caused.Given these problems,the multi-user task offloading method based on stochastic game is proposed.Firstly,based on the utility defined as the time saved by task offloading,the optimization model of utility for each RD is established.Secondly,through analyzing the randomness of tasks generated in each RD,the optimization model for each RD is transformed into the problem where each RD plays games for maximum utility.Then a stochastic game model is established and theε-Nash equilibrium solution is confirmed.After that the backward iterative algorithm is proposed to solve this problem.Finally,the strategies from the proposed method are compared with these from the static game,the simulation results show that the strategies from stochastic game can significantly reduce the execution time of the task and increase the efficiency of the mobile cloud.