基于马尔可夫相遇时间间隔预测的拥塞控制策略

Congestion control strategy based on Markov meeting time span prediction model

提出的基于马尔可夫相遇时间间隔预测的拥塞控制策略(Congestion control strategy based on Markov meeting time span prediction model,CCSMP,主要是通过规定节点缓存的排队方式和丢弃机制,将预测得到的较早与目的节点相遇的报文排于队首,尽可能丢弃效用值较低的报文,进而解决由于节点缓存有限而带来的拥塞问题。通过在ONE环境下进行仿真,与Drop-Front(DF)和Drop-Oldest(DO)两种拥塞控制策略对比表明:文中提出的拥塞控制策略提高了报文投递率,减小了平均网络时延,并且在一定程度上减少了网络负载比率和丢包率。

In order to solve the problem of the low delivery ratio caused by the limited resources in Delay- Tolerant Networking （DTN）, the routing mechanism based on multiple copies is usually used to improve the delivery ratio. However, it may lead to the problem of the overload of node buffer caused by the excessive numbers of copies, then network congestion happens. This paper proposes the congestion control strategy based on Markov meeting time span prediction model CCSMP. It mainly specifies the line method and discarding mechanism of nodes buffer. The predicted messages meeting earlier with destination node are lined in the first, and the messages with lower utility values are discarded as far as possible. Then the congestion problem caused by the limited node buffer is solved. Simulation in ONE environment is carried out to compare the proposed strategy with the two congestion control strategies, Drop-Front（DF） and Drop-Oldest（DO）. Simulation results show that the congestion control strategy proposed by this paper improves the message delivery ratio, reduces the average network delay, and to a certain extend reduces the overhead ratio and packet dropped rate.