The interconnected networks are facing with critical congestion issue due to the rapid growth of the traffic and the information on the links with limited capacity. We show that the traditional routing strategies are generally confronted with a tradeoff between the network capacity and the link usage when applying to the link-coupled network. To take a step to the issue, we propose a global dynamic routing (GDR) strategy that can simultaneously achieve multiple improvement of the transport performance with acceptable computational complexity at the cost of the average arrival time. Further analysis indicates the improvement is related to the nontrivial load distribution resulting from GDR. More surprisingly, the simulation experiments suggest our strategy GDR can suppress the occurrence of Braess-like paradox, which is a long-standing problem in transportation.