The current Internet suffers from high packet loss rates and large queueing delays. The oscillation of queue size at the congested link is an inherent problem of the network protocol and a main factor of degradation of network performance. This thesis proposes a new active queue management (AQM) algorithm to reduce the oscillation of queue size and alleviate network congestion.
The proposed AQM scheme controls flows using packet drop or ECN bit when congestion occurs. The congestion level is measured based on average queue size, average packet receiving rate and average packet drop rate. From those parameters, the proposed scheme calculates the drop probability so that the queue size is maintained stably. The drop of each arrived packet at the link is determined based on the calculated dropping probability.
We present performance evaluation of the proposed scheme through extensive simulations. The results show that the proposed scheme successfully reduces queue size oscillation and thus maintains high utilization while maintaining small queue length.