Abstract: Random exponential marking (REM) is an effective algorithm for active queue management (AQM), but suffers from some drawbacks such as slow convergence speed, and slow sluggish response to dynamic network traffic. Theoretical analysis shows that REM has similar property to PI scheme which could not detect congestion effectively. To overcome these shortcomings, an enhanced REM algorithm, EPREM, is proposed, in which an enhanced price is added and bandwidthbuffer proportion factor is introduced to tune the parameter λ. The simulation experiments based on NS2 show that EPREM can achieve faster convergence speed and better adaptive to variable network conditions.