The following text is copyright 2008 by Network World, permission is hearby given for reproduction, as long as attribution is given and this notice is included.
What a fragile and tangled communications web we have woven
By: Scott Bradner
A ubiquitous Internet requires ubiquitous connectivity. And we are doing quite well in this arena. Of course there are significant parts of the world and parts of the US where there is far from enough transmission capacity but those areas are shrinking everyday. What we do not have in too many places is the real redundancy that is required for high reliability.
This is the case for most enterprises and for too many countries. Few enterprises have set their networks to include redundant connections to sites or buildings, never mind redundancy to phone closets. Quite a few do have redundant Internet connections, but in many cases the redundant connections share common or adjacent physical facilities (e.g, fibers or conduits). Even where an enterprise has paid the local phone company for physically diverse redundant paths I know from personal experience that the phone technicians will come along at some point and "fix" the silly longer path so that it is shorter, faster and more efficient and, it just so happens, in the same fiber as the other link.
This lack of redundancy within enterprises and thus, a potential lack of reliability, have lead many telecom experts to conclude that VoIP will never be a major enterprise telephony solution. That conclusion mostly seems to come from old-line telephone folk who also think that VoIP will never be useful without central quality management. (I heard that argument 2 weeks ago.) But they tend to ignore the very high reliability of modern network equipment and the near ubiquity of cell phones. Even if your VoIP desk phone were to die you still can call on your cell phone. In this case the redundancy of two unrelated communications systems providing the same service (making phone calls) provides for almost perfect reliability.
What got me thinking about redundancy was the January 31st story that a good chunk of the Internet connectivity for the Middle East and South Asia died due to breaks in undersea cables. In this case there were two cables, run by different companies, but running under the Mediterranean Sea near each other. Something cut both cables, maybe a ships anchor dragging in rough seas. This was a case of theoretical but not quite real cable redundancy. As I write this, some connectivity is returning as ISPs switch to satellite paths. But, since 95% of all transoceanic and similar connectivity is through fiber cables it can be a big hit when a cable gets cut and worse when more than one is involved.
This is not the first time recently that something like this has hit Internet connectivity via undersea cables. A bit more than a year ago an underwater earthquake broke 9 submarine cables disrupting Internet service all over the Far East. (See http://www.iscpc.org/information/ICPC_Press_Release_Hengchun_Earthquake.pdf)
Clearly the Internet is important enough to international communications that more thought will have to be paid in the future to achieving real divergent path redundancy - for example linking the two cables to others to create rings around the globe - a break might make the traffic take the log way around but would keep things going.
In the old days, phone regulators would just say "do it." It would be done and the prices raised to pay for it - that is no longer the way things work -- much for the better in almost all cases, but maybe not in this one.
disclaimer: Harvard does not grok "just do it" - instead assumes 'proper discussion' is required for everything. In any case, the above discussion about redundancy represents my own opinion.