In this study, the authors propose a novel decentralised robust routing control strategy for average and covariance of queue length in a data network. The objective is to control the transitional behaviour of the data networks in response to new traffic load. For each node, the authors design a local state feedback controller based on minimising the worst case of the queue length. The authors then propose an innovative model for covariance of the queue length in each node. To control the average of queue length, local delay-dependent robust controllers are designed based on an equivalent descriptor representation and linear matrix inequalities (LMIs). For covariance control of the queue length, the authors first show that the covariance control problem is equivalent to a standard disturbance rejection problem. Then the authors propose a covariance feedback control law. To investigate the effectiveness of the proposed methodology an example is also presented and extensive simulation results are conducted. Simulation results indicate that the proposed method significantly improves the packet loss and system response time and is able to efficiently control the transient conditions in each.