A Neuro Fuzzy Approach for Performance Enhancement in TCP IP Differentiated Services for Congestion Control

Abstract

The problem of congestion is ubiquitous and is from the day of inception to till date. A number newlineof solutions have been proposed by various researchers and all are good or bad under different newlinecircumstances. In today scenario, especially after corona-19 pandemic and the usage of OTT newlineplatform, social media usage, and ChatGPT, the usage of internet has been raised exponentially newlineand hence the management of congestion has arisen with a new challenge. Thus, in a network, newlineespecially when the demand is of in real-time video and audio, the traffic needs to be managed newlinein a challenging manner. The problem becomes worse when the real-time video, audio traffic, newlineand other low priority traffic are on the same network. newlineFor such networks, congestion needs to be managed by minimizing, keeping in view of newlineminimal packet loss, minimum delay, minimum jitter, and maximum bandwidth utilization. newlineThe traditional approach to control congestion, of which mostly uses a linear regression newlinealgorithm for managing all types of networks irrespective of priority, which leads to slow down newlinetraffic. The traffic which needs to be prioritized in comparison with other traffic get the same newlinebehaviour which creates a problem in real time scenario. Thus, an algorithm needs to be newlinedeveloped which may forward or drop different packets in a different manner depending on newlinepriority. In a simpler way, a multilevel threshold value has been proposed for dropping the newlinepackets in case the maximum threshold value may be achieved. newlineThe Fuzzy method helps to create linguistic rules with multi-variable values while neural newlinemakes the system as adaptive learning while training only the centralized nodes. Thus, a model newlinehaving neuro-fuzzy tool will enhance the performance of the network, which gives priority to newlinereal-time packets by providing a different threshold value at the time of dropping of packets newlinewhich may lead to high utilization of available resources with minimum delay and loses. newlineThe research carried out presents a new tabular matrix system for active