 |
|
|
by Dr. Stephen Farrell
Sometimes standard networking technologies just cannot work due to the challenged communications context - rather than due to the usual operational failures. For example, an orbiter around Mars cannot make use of standard Internet protocols because of the latency involved - bounded by the light-trip time between Earth and Mars which varies between about 4 and 20 minutes. However, even on Earth, people are sometimes cut-off from all existing communications networks, for physical or cost reasons. To take an example, the reindeer herders involved in the N4C project may be 50km from any network infrastructure and generally cannot make use of satellite communications for cost reasons.
In such challenged environments however, delay- and disruption-tolerant networking (DTN) can still provide some level of connectivity, albeit not the type of Internet connectivity to which we have become more and more accustomed in the last decade. In the N4C case, data from remote reindeer herders may be transferred to the Internet via the helicopters that service the remote community's other needs. While this does provide connectivity of a sort, the latency involved is even worse than in the case of Mars with the equivalent round-trip time being measured in days or weeks.
It turns out that there are numerous other examples where we could make use of DTN in challenged environments, from military tactical networking to applications in developing regions, or, more generally, sparse sensor networks in challenging environments. So, beginning around 10 years ago, a bunch of engineers, mainly from NASA JPL, but together with other members of the Internet community (in particular Vint Cerf) began efforts to develop a suite of protocols that can operate in such challenged environments.
That work is today primarily carried out by the academic and government research communities, but with an open Internet Research Task Force group (the DTN research group, or DTNRG, http://www.dtnrg.org/ ) acting as a central point where protocol specifications can be agreed and which also allows the DTN community to meet regularly and share information about current developments.
To date, the DTNRG have developed and documented an architecture and a pair of protocols for use in challenged networks (see the bibliography below for details). The main protocol for DTNs is the so-called “bundle” protocol, that, as the name suggests, bundles up application data and control information into the same packets so that communications involving challenged areas don't require multiple round-trips before application data can be transferred (one of the main problems with the use of existing standard Internet protocols).
The other protocol is called the Licklider Transmission Protocol (LTP) and is intended mainly for use over very high latency links that are encountered in deep-space communications (though it may also have uses in terrestrial sensor networks).
Current work in the DTNRG is focused on addressing security and reliability issues with the bundle protocol and with improving our understanding of the deployment issues that arise when using DTN protocols in trials and in realistic networking scenarios. For example, finding out the names to use for the various entities in a DTN can be a challenge – if a node is physically moving then its name may sometimes match, and sometimes mismatch, the other names used in the currently local infrastructure. Indeed, this is one area where the N4C project is expected to be of general use to the DTN community since N4C will address many of these deployment issues and its solutions will hopefully prove to be broadly applicable.
DTNs are but one aspect of a broad set of somewhat connected efforts to engineer a future Internet that will be able to serve the expected needs of the many, many people and devices that will be communicating in the coming decades. In thinking about the future of DTNs we should first consider the scale of deployment of the current Internet protocol suite (TCP/IP) which is already deployed on billions of systems worldwide with billions of messages transferred daily and peta-bytes of (mainly user-generated) content stored. One possible outcome is that DTN might see wide-spread deployment as a standard part of most device network stacks, for use when those devices are used in challenged networking contexts. Another possibility is that DTN will only see deployments in niche areas, where challenged communications are a high-priority. The work of the DTNRG is certainly aimed at the former outcome – but, of course, predicting the future is a very uncertain exercise.
For the present, DTN is a technology that has reached a sufficient level of maturity (e.g., there are open-source reference implementations available), that we can begin modest deployments now, for situations where networking in challenging environments is a priority. From those modest deployments, (such as envisaged in N4C), we hope to better learn how DTN might fulfil the promise of becoming a standard and ubiquitous part of the future Internet.
Burleigh, S., Ramadas, M., and Farrell, S., “Licklider Transmission Protocol – Motivation”, Internet RFC 5325, September 2008.
Cerf. V., et al., “Delay-tolerant Networking Architecture”, Internet RFC 4838, April 2007.
Farrell, S., and Cahill, V., “Delay and Disruption Tolerant Networking”, ISBN 1-59693-063-2, Artech House, 2006.
Scott, K., and Burleigh, S., “Bundle Protocol Specification”, Internet RFC 5050, November 2007.
Dr. Stephen Farrell is a research fellow at Trinity College Dublin. Stephen researches and teaches mainly on delay tolerant networking and security. Prior to that, Stephen worked in Industry for 16 years including as a product architect for a Siemens subsidiary and as director of research for Baltimore Technologies. Together with Vinny Cahill Stephen recently wrote the first textbook on delay- and disruption-tolerant networking.
