The interplanetary network Vint Cerf envisioned years ago had its first real test recently.
The Epoxi spacecraft, which carried the Deep Impact probe to Comet Talent 1 in 2005, had its software reconfigured after delivering the payload to work as a test bed for Nasa's new Disruption-Tolerant Networking (DTN) protocol. As the craft dropped back towards Earth for one of the gravity assists that will ultimately sling it back towards the comet in 2010, it transmitted simulated images of the Martian moon Phobos using the new protocol.
The trial turned Epoxi into one of 10 nodes in a test network (the other nine were on Earth), to verify the reliability and robustness of the new networking architecture.
This new networking system, an outgrowth of Cerf's Interplanetary Net project, can be layered on top of TCP/IP, the protocol that today's internet uses. But although DTN is designed for a different environment than Earth, ultimately the technology may find its way back here, to improve communication back home.
How to network in space
JPL's Adrian Hooke, team lead and manager of space-networking architecture for Nasa, explained the limitations of TCP/IP-based internet. Although we tend to think of the internet as routing around faults, he said, it is "not actually tolerant of disconnection between two machines". If you lose a link between relay stations (routers), he explained, "the routers start dumping packets on the floor after a few milliseconds".
Out in the solar system, where distance means point-to-point communication time of a single bit can take minutes or hours, and where there is no system of interconnected routers, relay stations need to be smarter and more robust. Dropping packets doesn't work. "In space, it's very rare that you have an end-to-end path," Hooke said.
DTN devices don't just send off packets to the next device in the communications chain, as routers do. Instead, they hold on to packets until they expect that they will be received and, after they send them, they keep holding on to them until they receive an acknowledgment. Only once the packets are acknowledged do they release 'custody' of the data to the next link in the communications chain.
DTN networks need more smarts and storage than typical routers. They need to know which devices they can send to, and when, since planets and space vehicles don't stay put. And they need enough storage to hang on to packets that are coming in even when there may not be a receiver onto which they can offload them.
These concepts are not new. Email routers use store-and-forward architectures to transfer information, and mesh networks are opportunistic with their connections. But getting the DTN protocols certified for space operations requires a lengthy development cycle. Hooke said Nasa hopes to have DTN ready to be built into spacecraft and ground-based radios in 2011, but that it will be four or five years after that before the technology will then make it into space. In 2015 or 2016, he said, "an interesting cluster of missions to the moon" will be launching, and he hopes to see DTN on them.
He also expects DTN to be part of the communications protocol for the Mars Sample Return mission, which is scheduled to launch in 2020 (but will probably be more like 2025). In that mission, a "full fleet of spacecraft" from several countries will all need to interoperate, and DTN should make the communications more reliable, and easier to build, than a patchwork of point-to-point radios. But first Nasa and other space agencies need to know it works.
DTN back on Earth
DTN concepts are being applied to similarly flaky networks back home. Not surprisingly for a Darpa-funded project, the US military is experimenting with DTN for "stressed tactical military communications". On a battlefield, as in space, there's rarely an existing communications infrastructure a device can drop in to, so data radios need to be more tolerant of poor networks and opportunistically take advantage of communications links when they are available. Likewise, the US Navy is looking at DTN to help submarines send and receive data in bursts when they surface or come close to a relay buoy.
DTN can integrate with existing TCP/IP networks, Hooke said. 'Bundle agents' can sit on the internet and handle the store-and-forward protocols as well as the transfer of data from occasionally-connected devices to the main internet.
Not all applications are military: a team in Sweden is using DTN to track reindeer movement (via geolocators tagged to animals), for example. Intel is looking at DTN to build out networks in developing countries with no communications grid. And in the US, cellular equipment manufacturers are thinking about DTN for devices at the edges of expanding networks.
