Accelerating the Network Congestion Control in Classic Kelly Model Implemented by 2 ECN Bits
Weirong Liu, Min Wu, Jun Peng, Guojun Wang · 2009
To get the global fairness in Internet, the Classic Kelly Model proposed by F.P.Kelly uses the utility function to optimize the data fluid control. The model gives two algorithms to control the data send rate. One is prime algorithm, and the other is dual algorithm. The two algorithms can converge to proportional global fair point. However, the two algorithms use only the positive price to feedback the congestion information in network link, which results in the slow convergence rate. In addition, many mechanisms based on the two algorithms need to introduce extra packet header to get price information. To solve the problem, this paper presents a 2 bits ECN mechanism to accelerate the model convergence rate. In the innovate mechanism, the link price can be negative and the price can be conveyed by using a new ECN bit without adding extra packet or packet header, means that, 2 bits congestion transporting mechanism is enough. By using the 2 bits, fairness can be keep and asymptotic stability can be reached rapidly.