FLoadAutoRED: an active queue management scheme to prevent congestion in a dynamically varying traffic in IP networks

Abstract

The development and need of distributed data and multimedia applications are growing rapidly in Internet. Some common examples are Video conferencing, Internet telephony, File transfer, On-line games; Internet based banking, Email distributions, News updates and E- business. High quality delivery of data requires high network bandwidth. Moreover, the best quality of service is required to communicate the desired information across the IP network for these application-oriented operations. The special characteristics of these multimedia and data applications place a number of requirements on the IP network. The requirements are in terms of throughput, packet loss, and delay. Further, the deployment of these applications increases the percentage of dynamic varying traffic in the Internet resulting in congestion and starvation of TCP traffic. Such an effect in Internet results in multiple packet losses, low throughput and low delay leading to a congestion collapse. Internet multimedia and data applications cannot tolerate such performance degradation due to congestion. Congestion has to be handled efficiently that keeps the IP network in a stable state to accommodate varying traffic with high performance. Hence congestion has to be prevented by implementing a proactive approach i.e Congestion Preventive technique that keeps the IP networks in a stable state to accommodate varying traffic.In IP networks, congestion can be detected and prevented at the network level (i.e.) at the Router level. Each router uses a finite buffer maintained in a queue fashion for the packets to be stored and transmitted. As a result, in case of heavy traffic, router gets congested and further congestion collapse is likely to occur. Packet drops at the routers implicitly indicate congestion to the end sources in the network for it to adjust accordingly the sending rate or window size.Now-a-days Internet applications with dynamically varying load floods the Internet routers with data that requires efficient Queue management mechanisms.