When I got my first class on computer networks, I was hooked. I loved the idea of small packets of information hopping from one computer to another. Amazed about how the information always seemed to arrive at the right spot, even if there were a gazillion computers connected, like on the Internet. Although I never worked in that particular field of information technology, I still remember an important lesson from the routing-algorithms that could be used. To find out which way another computer was located within the network, you can use one single computer as the main guide; that host has all the data needed to locate the computer you want to send your information package to.
This sound very effective at first, and it even is, if not too many PCs and mainframes are connected to the network. However, when you are thinking about the Internet, forget it. The information is just too much, and always outdated if you try to have a single map of the net. You have also a single point of failure in this scenario. If this one computer crashes, not one package will arrive at its destination. In search for alternatives, my mind was fixed on needing a map of the network. As it turned out, you can also have algorithms without the need for a image of the entire network; if you get a data package, you just give it to a computer you are connected with, and that accepts it the fastest. I never forget the new: the hot-potato-algorithm. After a while, the package will end if with its destination.
(more…)