Performance Evaluation For DTN Routing Algorithm Based On Replication Strategy In Real Vehicle-Route Scene

With network technologies updating and advancing, the network applications becomeflourishing. Classical internet architecture has been unable to meet the need of some areas,such as Space Exploration, Battlefield Information Acquisition, Water Quality Testing, andRemote Communication and so on.With the same basic features of this kind of area: large delay exists in the network, thenetwork topology changes frequently and periodically, the transmission rate is not equal toeach other between nodes, high error rate exists and node’s capacity is limited. Delay TolerantNetwork (DTN) is a new type of network emerges as the times’ requirement to solving aboveproblems. In purpose of applying DTN more efficiently in more areas, the way to evaluatingrouting algorithm appropriately and objectively is essential.In existing routing protocols and algorithms for DTN, some important are based onreplication strategy: the Epdemic and Spray And Wait algorithm are based on Flood theory;the PROPHET algorithm is based on the probability theory; the MAXPRO algorithm is basedon scheduling theory. All of above are mature algorithms.According to these replication strategy DTN routing algorithms, we have designedsome experiments to justify the status of these algorithms and to show their performance,using the opportunistic network environment simulator—ONE. Innovatively, we adoptedONE’s open model which called External Movement, with the data of real vehicle trace inreal world. We have completed155tests, by4kinds of parameters changed. Two groups ofthe original data we adopted in the tests derived from the objective world: one is more than500taxies’ trace data in San Francisco; the other is more than7000taxies’ trace data in asouthern city of China. Of course, we have sampled and normalized these data. Because thetaxies are similar to other small capacity vehicles in the way of activity and running trace,they can fully represent the general vehicle activity rules. Because of the data adopted rootingfrom real world, the experiment largely eliminated the limitation of simulation of DTNnetwork movement model which based on software. It reflected the characteristics moreaccurately in the work of the routing algorithm. Finally, we deeply compared the5algorithms by3indexes: the delivery rate, theaverage delay time and the cost of resources and draw the general conclusions.