A Networking Idea That’s Out of This World


Millions of miles above us, a new networking technology is taking shape that could one day help improve how applications are networked on Earth.

NASA is testing a network layer technology that can withstand the rigorous demands of space communication better than the standard TCP/IP protocol, which dominates terrestrial (non-space) networking technologies.

Officially called Disruption Tolerant Networking (DTN), the technology went through testing earlier this month with a space probe that is currently more than 20 million miles from Earth.

DTN uses a different kind of approach than TCP/IP for packet delivery that is less cumbersome and more resilient to disruption than TCP/IP.

“In fact, far more research has been done to date on the application of DTN to terrestrial communication problems than on its use in space flight missions,” said Scott Burleigh, a senior engineer for the Deep Impact Networking Experiment of NASA’s JPL (Jet Propulsion Lab). “DTN has potential benefits in providing connectivity to parts of the world that are under-served by existing network infrastructure, in supporting oceanographic research, in tactical military communications, and more,” he told InternetNews.com. “It’s a pretty active field.”

The basic idea behind DTN network endpoints aren’t always continuously connected. In order to facilitate data transfer, DTN uses a store-and-forward approach across routers that is more disruption-tolerant than TCP/IP. However, the DTN approach doesn’t necessarily mean that all DTN routers on a network would require large storage capacity in order to maintain end-to-end data integrity.

“It’s always possible to have a DTN router that happens to be in constant communication with all of its neighbors over links on which round-trip times are very short, in which case very little storage would be needed,” Burleigh explained. “All the bundles it received would immediately be forwarded, much as in an Internet router.”

If a DTN router is required to hold onto data for some length of time, then it would need some place to park that data. The amount of storage needed would depend on the difference between the incoming and outgoing data rates and the maximum length of time that the former may exceed the latter.

In terms of network addressing for DTN, NASA is using an overlay technology that could include IPv4 or IPv6 address spaces. Burleigh explained that DTN is based on a new protocol named Bundle Protocol (BP; RFC 5050). Bundle protocol operates as an overlay protocol that links together multiple subnets (such as Ethernet-based LANs) into a single network.

“BP is an overlay protocol that can link together multiple networks, some of which might be IPv4-based or IPv6-based, but some of which might be much less familiar networks such as sets of deep space links,” Burleigh said.

The regular Internet uses DNS (domain name servers) to connect domain names to IP addresses. For now there is no similar DNS system for space-based traffic. Burleigh noted that DNS is not an issue for NASA right now since the number of different devices isn’t yet so large that referring to them by number rather than by name is a problem. For potential non-space DTN applications Burleigh claimed the current Internet DNS could be used.

“Eventually there may additionally be a requirement for some sort of native DTN endpoint ID directory service, but we don’t yet know exactly what that service would have to do or how it would have to function,” Burleigh said.

NASA already has its own Deep Space Network for communicating with its fleet of spacecraft spread across our solar system. With DTN, the core technologies that NASA is currently using at the lower levels of the communications stack will remain the same.

Burleigh noted that at the physical data link layer NASA is still using radio transmission and reception via the big tracking stations of the Deep Space Network. The difference with DTN is at higher layers.

“This is the first time we’ve operated a network-layer protocol on a deep-space spacecraft, so it’s the first time we’ve been able to use the spacecraft as an automated router in deep space,” Burleigh said. “The Mars Exploration Rovers send data to Earth by relay through the Odyssey orbiter, but that relay operation is all manually commanded. The DTN software would do the same thing automatically and would at the same time be more flexible and efficient.”

Next page: An End to Chatty, Earthbound Networks?

Page 2 of 2

An End to Chatty, Earthbound Networks?

Here on Earth, enterprises are spending hundreds of millions on WAN optimization technologies designed to get over the limitations of TCP/IP. The general argument is that TCP/IP can be excessively chatty, and using up more bandwidth in the process, for example. Burleigh said DTN and TCP/IP are optimized for different parameters.

“It is true that DTN is designed to be much less ‘chatty’ than TCP/IP Burleigh explained. “The origin of the term ‘bundle’ — which you can think of as the DTN equivalent for ‘packet’ — lies in the notion that some metadata may accompany the data, so that the receiver doesn’t have to ask the sender for additional information in order to know what to do with each bundle when it arrives.”

That said, DTN does have its own bandwidth costs. DTN may reduce end-to-end data delivery latency by sending a little extra data, but that extra data costs bandwidth.

While NASA is testing DTN millions of miles above the Earth, you can test it out here, too, via an open source implementation of DTN available from SourceForge. It’s not the same implementation that NASA tested on its spacecraft. The exact implementation that NASA used is available for download from https://ion.ocp.ohiou.edu.

The DTN effort is cross-platform and operating system independent. Burleigh said the DTN implementation that NASA has tested on its spacecraft is written in C. It is designed to run in multiple operating environments without any adaptation or modification including VxWorks, Solaris, OS/X, several different distributions of Linux, and others.

“The intent is to be able to use the same software both in flight and on the ground so that any enhancement or bug fix we add only has to be implemented once, meaning we don’t have to deal with differences in the way DTN works on different platforms,” Burleigh stated. “We hope this approach will limit costs and make the system continuously more robust and capable over time as its installed base grows.”

News Around the Web