Abstract: In order to mitigate the switch congestion caused by bust data in the network, a preventive mechanism targeting for the data storm in switch is proposed. Firstly, it promotes the number of tokens in a token bucket with an adaptive rate, and simultaneously monitors the type of data frames from a port based on user software configuration. Then, the number of token required by forwarding that specific type data frame is calculated. Finally, the number of token in the token bucket and required by forwarding data frame is compared. Further, if the token amount in bucket is more than that required by data frame forwarding, the proposed mechanism permits data frame forwarding. Otherwise, the flow control is activated. In the realistic application, the collaborative work of multi-token is used to implement the flow control for multiple kinds of data frame. Simulation results show that the proposed mechanism, based on multi-token technique, can successfully suppress the data storm inducing by unicast, multicast and broadcast. All that can completely satisfy the project's requirements.