Modeling Dynamic Patterns Adapted Joint Multidimension Resource Scheduling via Graph Sequence in Optical Data Center Network

Traffic patterns in data center network (DCN) may have distinguished features in a graph view. The traffic patterns may vary by time in DCN, and network topology reconstruction may help to adapt these dynamic traffic patterns. Traditional network scheduling may ignore the feature of patterns, which results in performance deteriorations. Optical DCN with multidimension resources may provide topology flexibility. If the multidimension resource scheduling problem is modeled as topology reconstruction by time, it may benefit the dynamic traffic patterns. To achieve this goal, this paper used a graph sequence to represent the time-varying network topology, and then a graph sequence based scheduling (GSS) algorithm has been proposed to optimize the traffic patterns. By simulation on our network modeling tool, the effectiveness on lowering latency and maximizing the link utilization of GSS has been verified.