| ¡¡ | Chinese Journal of Computers Full Text |
| Title | Analysis on Congestion Control Strategy Based on Time Series Deduction |
| Authors | XU Li-Bo WU Guo-Xin |
| Address | (Key Laboratory of Computer Network and Information Integration, Southeast University, Nanjing 210096) |
| Year | 2007 |
| Issue | No.9(1638¡ª1644) |
| Abstract & Background | Abstract Congestion control is one of key technologies in high-speed networks. Through building accurate Markov chain network models for TD and RED arithmetics, the performances of various congestion control strategies are deeply studied and compared by theoretical analysis and numerical simulation. In particular the mathematics process and theoretic source of average queue size are analyzed from time series measures. Then a new calculating formula based on time series deduction and its network model are given. The results show that new arithmetic brings obvious advantages on reliability and resource utilization. Its trend is more close to ideal value curve. At last the performances of arithmetics with various priorities discarding strategy are reviewed. keywords congestion control; average queue size; time series; Markov chain; EWMA background To avoiding congestion collapse on Internet links, relay devices such as routers adopt AQM to tolerate bursting flows. IETF introduces RED to design AQM arithmetic.To combing end-user activity to RED,there appear several new AQM schemes to improve RED control ability. For no modeling expression, majority of these schemes lack theoretic explication and deduction. The main contribution of this paper is modeling and analyzing various AQM strategies. Through Markov chain description, the model of RED behaves under unresponsive and responsive end-flows is built. Then the nature and expression of average queue length in RED are observed. At last a new arithmetic with time series ARMA for forecasting average queue length is given and the simulation results show its obvious advantages on reliability and resource utilization. |