Congestion control in disruption-tolerant networks: A comparative study for interplanetary and terrestrial networking applications

Controlling congestion is critical to ensure adequate network operation and performance. That is especially the case in networks operating in challenged- or extreme environments where episodic connectivity is part of the network’s normal operation. Consequently, the “pure” end-to-end congestion control model employed by the Internet is not adequate. Our goal is to study congestion control mechanisms that have been proposed for these so-called disruption tolerant networks, or DTNs. In this paper, we conduct a performance study comparing existing DTN congestion control mechanisms for two main application domains, namely: inter-planetary (IPN) and terrestrial networking applications. Our results confirm that congestion control helps increase message delivery ratio, even in highly congested network scenarios. Furthermore, the results show that existing DTN congestion control mechanisms do not perform well in IPN scenarios. Our study also suggests that good design principles for congestion control in DTN scenarios include: combining reactive and proactive control, using local information instead of global knowledge, and employing mechanisms that are routing protocol independent. One important conclusion from our quantitative study is that there is currently no universal congestion control mechanism that fits all DTN scenarios and applications.