Abstract: Currently, traffic engineering is typically deployed to improve the utilization of data centers (DC) backbone networks, which usually belongs to the same online service providers. Although the efficiency is remarkable, the bandwidth allocation fairness of different aggregate flow isn’t considered. Hence, the QoS guarantee is restricted. Because the bandwidth resource is expensive and packet loss is typically thought unacceptable, the bandwidth utilization should be maximized, at the same time, the QoS guarantee of different flow should be improved. In this paper, the problem of contending the share bandwidth is modeled as a cooperative game, and different aggregate flow compets the share bandwidth and maximizes the overall bandwidth resource utilization simultaneously, and the optimal bandwidth allocation policy, called cooperation game based bandwidth allocation (CGBA), is obtained through searching the Nash bargaining solution (NBS) of the game and balancing the tradeoff between minimum bandwidth guarantee and bandwidth allocation fairness. Simulation on a Mininet testbed shows that the proposed policy can effectively guarantee minimum bandwidth of each aggregate flow while ensuring the allocation fairness, compared with three other classical bandwidth allocation policies.