Abstract: In this paper, an additive-increase and multiplicative-decrease （AIMD（a,b）） algorithm is first outlined, and then a dual AIMD-based TCP congestion control （D AIMD） is presented, whose implementation is still based on AIMD（a,b） algorithm except that parameter a and b is variable. The main idea of D AIMD is that AIMD（a,b） algorithm for small a and b is used in the steady state while AIMD（a,b） algorithm for large a and b is used in the transient state. Theory analysis and simulation results show that D AIMD has the following characteristics: ① it can be easily implemented with lower additional overheads; ② in the transient state, network available resources can be quickly utilized while network heavy congestion can also be swiftly alleviated; ③ in the steady state, the fluctuation of data sending rate can be mitigated which results in higher utility of network resources; ④ D AIMD flow and TCP flow can fairly compete network resources with each other.