An underlying principle of the Internet is “global reachability”–the concept that every Internet address should be addressable from any other address [1]. When we fail to reach a certain address, the first instinct is to blame the local network connection or the system serving that individual page.

But could there be a more fundamental reason? Researchers at the University of Washington believe that there is. Katz-Bassett et. al. have published a paper [1] that aims to identify long-standing problems with reachability and to provide information to network operators about how to correct these problems. Their tool, Hubble, also report real time, overall statistics and presents the information about reachability problems on a world map at

http://hubble.cs.washington.edu/

The following example of Internet problems identified by Hubble is included in a MSNBC article [2] describing the work:

The motivation of this work is the unreliability of the Internet: the authors argue that their results show that the Internet is much less effective at delivering information than the telephone network. A very interesting consideration to make for this problem is the impact of network effects. The Internet has been modeled as a graph [3], and it’s reasonable to consider that problems in one area will lead to issues in other areas. Persistent failures and interconnectedness are likely at the heart of the black holes problems in the aforementioned investigation. Although Internet routers are not people, many network effect concepts may be applicable. For example, if one router is not available, this may increase the load on another router, causing it too to collapse and to cause the start of what may become a widespread “black hole” of unreachable addresses. The unreachable addresses may have nothing to do with the routers that failed; it may only be the path between the two nodes that is effected.

The University of Washington researchers suggest several topologies that illustrate the forms of problems that lead to the Internet black holes that they are seeing using their monitoring tool. Interpreted as a graph, it may be possible to extend this work with network theory.

In the future, it would be interesting to work modeling computer networks as complex graphs [4], with work examining the technical causes of failures in the network. Theoretical studies in networks may shed light on how to correct problems with the Internet. On the other hand, a deep understanding  of the technical aspects of these problems will lead to a better understanding of theoretical information networks.

[1] http://www.cs.washington.edu/homes/ethan/hubble/nsdi08/hubble.html

[2] http://www.msnbc.msn.com/id/24067737/

[3] http://www.caida.org/research/topology/as_core_network/
[4] http://www.caida.org/publications/papers/2008/self_similarity/