Performance enhancement of AIMD algorithm for congestion avoidance and control

• Main Author: Jasem, Hayder Natiq
• Format: Thesis
• Language: English
• Published: 2011
• Online Access: http://psasir.upm.edu.my/id/eprint/30923/19/FSKTM%202011%2034R.pdf; http://psasir.upm.edu.my/id/eprint/30923/

Computer networks have revolutionized the human life in the end of the last century. Two major technical differences between telephone communication and the computer networks are as follows. First, telephones use a dedicated physical wire for each communication between two parties (circuit switched network). In contrast, computers share this physical line between many – unlimited, at least theoretically – computers (packet switch network). Secondly, telephone communication is used only for voice messages while the packets in packet switch network can contain any kind of information, even voice messages. It should be obvious that these automatic robots (computers) need specific rules and regulations (protocols) to be able to communicate, interleaving data, through a single connection channel (e.g. bottleneck link). These protocols have been developed and improved since the advent of the Internet. One major issue arose in 80s when routers got congested and the connections collapsed because they were pushed to their limits by the competing robots for connection. Since then, many mechanisms have been suggested and developed to solve this problem. New applications in computers and their specific requirements for communication have even added more complexity to this issue. Hence, more work needs to be done. Additive Increase Multiplicative decrease (AIMD) is one of the best principles offered to solve this problem. This study is also in this field and offers an enhanced version of this mechanism. The contribution of this research is to improve the performance of AIMD in terms of efficiency (the network utilization) and fairness. National Chiao Tung University’s network simulation (NCTUns) has been used in this development to compare the new algorithm with the older versions and determine its advantages over the older versions. In addition to theoretical (mathematical) proof and to provide laboratory confirmation,this research has also used NCTUns to conduct tests and experiments to measure different performance metrics of the new algorithm such as end-to-end delay, link utilization, efficiency and fairness, and compared the results with the results of the same tests on older versions. The results have shown improvement in all aspects particularly the major ones, efficiency and fairness over the previous algorithms.