Queue Assisted Random Early Detection for Congestion in Software Defined Networks

Software Defined Networks (SDN) utilize the network devices as virtual devices to analyze the traffic and transmit the relevant packets. It is quite beneficial for the internet of Things (IoT) where the huge number of devices share the packets which cannot be handled by the conventional networks. SDN is also essential to manage congestion scenarios during massive communication especially in video streaming. The main problem arises when the residual queues on the network path are left underutilized whereas the bandwidth is reduced as per the minimum bandwidth link between any two nodes in the path. It occurs based on the feedback based congestion avoidance mechanism. This work presents a proposed mechanism to utilize the residual queue in the communication path to enhance the minimum bandwidth. It maintains the average queue length to analyze the percentage of queue utilized and then decide about the packet dropping probability. In proposed scheme, the congestion can be managed earlier whereas the base scheme applies a constraint of double the size of previous threshold which may result in identifying the congestion at a later stage where the retreat is hardly possible. According to the findings in results after simulations, the proposed IAGRED dominates the counterparts in terms of queue delay, queue length, link utilization and throughput.