Even the dolphins use pocket switched networks!

Email, texting, Instant Messaging, Instant response...one of the things about modern telecoms is that they fuel our desire to "talk" to people anytime, anywhere, instantly. The old kind of mail is dismissed as "snail mail". A slow network is a frustrating network. So why would anyone be remotely interested in doing research into slow networks? Professor Jon Crowcroft of the University of Cambridge and his team are, and his kind of network could be the network of the future. The idea is already being used by the dolphins (not so surprising I suppose given according to Douglas Adams' "The HitchHiker's Guide to the Galaxy" they are the second most intelligent species on Earth...after the mice).

From node to node

Traditional networks rely on having lots of fixed network "nodes" with lots of fast links between them. These network nodes are just the computers that pass on the messages from one to the other until the messages reach their destinations. If one computer in the network fails, it doesn't matter too much because there are enough connections for the messages to be sent a different way.

There are some situations where it is impractical to set up a network like this though: in outer space for example. The distances are so far that messages will take a long time - even light can only go so fast! Places like the Arctic Circle are another problem: vast areas with few people. Similarly, it's a problem under the sea. Signals don't carry very well through water so messages, if they arrive at all, can be muddled. After major disasters like Hurricane Katrina or a Tsunami there are also likely to be problems.

It is because of situations like these that computer scientists started thinking about "DNTs". The acronym can mean several similar things: Delay Tolerant Networks (like in space the network needs to cope with everything being slow), Disruption Tolerant Networks (like in the deep sea where the links may come and go) or Disaster tolerant networks (like a Tsunami where lots of the network goes down at once). To design networks that work well in these situations you need to think in a different way. When you also take into account that computers have gone mobile - they no longer just sit on desks but are in our pockets or handbags, this leads to the idea of a "ferrying network" or as Jon Crowcroft calls them: "Pocket Switched Network". The idea is to use the moving pocket computers to make up a completely new kind of network, where some of the time messages move around because the computers carrying them are moving themselves, not because the message itself is moving. As they move around they pass near other computers and can exchange messages, carrying a message on for someone else until it is near another computer it can jump to.

From Skidoo to you

How might such networks be useful in reality? Well one already exists for the reindeer farmers in the Arctic Circle. They roam vast icy wastelands on skidoos, following their reindeer. They are very isolated. There are no cell phone masts or internet nodes and for long periods they do not meet other people at all. The area is also too large to set up a traditional network cheaply. How can they communicate with others?

They have set up a form of pocket switched network. Each carries a laptop on their skidoo. There is also a series of computers sitting in tarns spread around the icy landscape. When the reindeer farmers want a service, like delivering a message, the laptop stores the request until they pass within range of one of the other computers perhaps on someone else's skidoo. The computer then automatically passes the message on. The new laptop takes the message with it and might later pass a tarn, where the message hops again then waits till someone else passes by heading in the right direction. Eventually it makes a hop to a computer that passes within range of a network point connected to the Internet. It may take a while but the mail eventually gets through - and much faster than waiting for the farmer to be back in net contact directly.

Chatting with Dolphins

Even the dolphins are in on the act. US scientists wanted to monitor coastal water quality. They hit on the idea of strapping sensors onto dolphins that measure the quality wherever they go. Only problem is dolphins spend a lot of time in deep ocean where the results can't easily be sent back. The solution? Give them a normal (well dolphin adapted) cell phone. Their phone stores the results until it is in range of their service provider off the coast. By putting a receiver in the bays the dolphins return to most frequently, they can call home to pass on the data whenever there. The researchers encountered an unexpected problem though. The dolphin's memory cards kept inexplicably filling up. Eventually they realised this was because the dolphins kept taking trips across the Atlantic where they came in range of the European cell networks. The European telecom companies, being a friendly bunch, sent lots of text messages welcoming these newly appeared phones to their network. The memory cards were being clogged up with "Hellos"!

The Cambridge team are investigating how similar networks might best be set up and used for people on the move, even in busy urban environments. To this end they have designed a pocket switched network called Haggle. Using networks like Haggle, it might be possible to have peer-to-peer style networks that side-step the commercial networks. If enough people join in then messages could just hop from phone to phone, using bluetooth links say, as they passed near each other. They might eventually get to the destination without using any long distance carriers at all.

The more the merrier

With a normal network, as more people join the network it clogs up as they all try to use the same links to send messages at the same time. Some fundamental theoretical results have shown that with a pocket switched network, the capacity of the network can actually go up as more people join - because of the way the movement of the people constantly make new links.

Pocket switched networks are a bit like gases - the nodes of the network are like gas molecules constantly moving around. A traditional network is like a solid - all the molecules, and so nodes, are stationary. As more people join a gaseous network it becomes more like a liquid, with nodes still moving but bumping into other nodes more often. The Cambridge team are exploring the benefits of networks that can automatically adapt in this way to fit the circumstances: making phase transitions just like water boiling or freezing.

One of the important things to understand to design such a network is how people pass others during a typical day. Are all people the same when it comes to how many people they meet in a day? Or are there some people that are much more valuable as carriers of messages. If so those are the people the messages need to get to to get to the destination the fastest!

To get some hard data Jon and his students have been handing out phones. In one study a student handed out adapted phones at random on a Hong Kong street, asking that they be returned a fixed time later. The phones recorded how often they "met" each other before being returned. In another similar experiment the phones were given out to a large number of Cambridge students to track their interactions. This and other research shows that to make a pocket switched network work well, there are some special people you need to get the messages to! Some people meet the same people over and over, and very few others. They are "cliquey" people. Other more "special" people regularly cross between cliques - the ideal people to take messages across groups. Social Anthropology results suggest there are also some unusual people who rather than just networking with a few people, have thousands of contacts. Again those people would become important message carriers.

So the dolphins may be the "early adopters" of pocket switched networks but humans may be not far behind. If so it could completely change the way the telecom industry works...and if we (or the dolphins) ever do decide to head en-mass for the far reaches of the solar system, pocket switched networks like Haggle will really come into their own.

This article is based on a talk given by Jon Crowcroft at Queen Mary in January 2007.