Communications in deep space environment are facing many problems,including long progagation delay,weak signal strength,large bidirectional data-rate asymmetries,and arbitrarily long periods of link disconnection.Delay tolerate network (DTN) can well solve the communication problems in deep space mentioned above.The concept of DTN is abstracted from many new emerging network instances,which experience frequent long-duration partitions.This particular topology feature makes DTN's routing mechanism quite different from typical Internet,in which the existence of a path from source to destination is always guaranteed. In this paper, firstly we discuss the DTN system structure and Internet limitations. Secondly, we present a mathematical model for the one-hop delay which is the most important parameter in routing information set in DTN.After studying the delay in DTN from the queuing theory point of view,a delay model of DTN is constructed based on G-limited vacation queuing model.The delay parameter,which is the most important one in DTN model is extracted and then the performance analyzing is given. Thirdly, we proposed a routing algorithm based on Markov Decision,named MRDDTN,is proposed. MRDDTN is self-learning and can solve problems in complex network environment characterized by very long delay paths and frequent network partitions. Through simulative analysis, MRDDTN is found to be able to achieve better performance than Epidemic and PROHET routing algorithms under the same network conditions. |