Application-aware multi-tenant scheduling mechanism model of power cloud platform

Power multi-tenant data centers widely use distributed parallel computing frameworks to process data, and applications from different tenants share and compete for bandwidth resources. The distributed parallel computing framework mostly uses Coflow as the traffic model, and optimizing the Coflow scheduling can improve the performance of the distributed parallel computing framework. Aiming at the problem of performance optimization and tenant isolation during Coflow scheduling, an application-aware multi-tenant Coflow scheduling mechanism, Aamt, is proposed. First, according to the tenant isolation problem, the progress is introduced as a measurement standard; according to the dynamic and elastic changes of tenant demand, the tenant arrival model is designed according to the NHPP modeling, and the long-term progress of tenant isolation guarantee is obtained to achieve application awareness; in order to effectively utilize the network bandwidth, Lagrange is designed Dual optimization, modeling from two aspects of bandwidth allocation and completion time, to obtain performance optimization progress; experimental results show that Aamt can not only ensure tenant isolation, but also increase bandwidth and improve link utilization.