PMCC: A probabilistic model for congestion control (dealing with needless retransmission and waiting)

The aim of the TCP/IP protocols and various algorithms is to provide an end-to-end reliable communication in the World Wide Web communication with prime focus on quality of services. And in order to provide this the network which governs the communication must be congestion free. Congestion can be defined as: Any such event that consequence the degradation of performance and denial of services is congestion which are promised earlier (goal of reliable networking). The researchers are working for solution to the congestion problem from 1988. Many algorithms in this category have already been proposed but the dynamic and unpredictable nature of the network needs an adaptive algorithm in every aspect. Ones the congestion occur normally algorithms reduce sending rate and later on recover it, but The cost of facing congestion is very high. And since TCP is widely accepted protocol for reliable communication, hence any algorithm in this category must not violate the TCP semantics. So, algorithms must prevent congestion and detection and recovery must be the last option. In this paper, we proposed a probabilistic model for improving the performance of the network and try to prevent congestion with cubic window adjustment function and adapt the variability in the network automatically. Since we waste the maximum time during communication in waiting for the acknowledgement and in retransmission of lost packets, this algorithm mainly focused on the avoiding the unnecessary “waiting and retransmission” and minimizing the complexity to calculation of the congestion window size.