Casey Johnston at Ars Technica has a detailed write up of some new research published in Nature Communications this week. It is comparable to other efforts about which I’ve read before. The idea makes a good deal of sense, mapping out the existing internet backbone and looking for schemes to re-arrange the connections and routing for greater efficiency. In this instance, the mapped view is hyperbolic and relies on something called “greedy forwarding”.
Greedy forwarding is based on the desire to get information around in as short a distance as possible, while forcing each node to maintain very little information—nothing beyond the coordinates of its neighbors. If a node holds information, it checks the destination coordinate against those of its neighbors, and then sends the info on to the neighbor that is nearest to the destination. After a series of similar transitions, the information arrives without anyone having to chart the whole route.
The hitch is that the scheme doesn’t work with the way the various players on the backbone are already connected. Peering and transit relationships are driven more by business needs than routing efficiency. The likelihood of re-organizing for potential efficiency and reliability considering existing investments is slim.
Johnston does point to some cause for optimism, that the researchers are looking for a similar approach that might work despite the existing business connections. Calling the idea of a pipe dream may be a bit of an exaggeration after all, depending on those future efforts.