One of the things I most enjoy about the TERENA conference is hearing about new applications of computers and networks. My favourite so far this year was presented by Cees de Laat in his plenary talk yesterday: understanding when the dykes that keep most of the Netherlands dry are in danger of breaking. It turns out that earth dykes are rather complicated structures, needing just the right amount of moisture – too wet and they turn to mud and collapse, too dry and they crack.
Researchers have built an experimental dyke, with sensors for sound, pressure and other measurements on and within the structure. This can be exposed to unusual conditions and even (unlike a real dyke!) deliberately made to collapse. Information from sensors during these experiments can be used both to create better numerical models and to provide indications of when real dykes may be approaching a dangerous state. A particularly good measure is provided by the response to the tide rising and falling: the dyke is supposed to move under the weight of all that water, but a change in the normal regular pattern of movement may well be a warning of a problem.
To provide better warnings, there is now a desire to instrument all of the dykes in the country: the Internet suggests that’s several hundred miles. Although each individual sensor only produces a small quantity of data, the total bandwidth from a huge number of sensors is challenging. Depending on how the measurements will be used, combining them all together in one place may not be the most useful thing to do, since the meaning of one set of measurements may depend on other sensors from nearby, but be less relevant elsewhere in the country. This suggests that right network architecture may vary depending on the study being carried out – Cees’s research team are therefore looking at how networks programmable at a very low level may be used to incorporate networks, sensors, CPUs and storage into a large, dynamic virtual machine.