On Mon, 21 Apr 2003, Steve Gibbard wrote:
This is NANOG, and this is pretty basic, so this is probably the wrong forum for this explanation. That said, if a small ISP gets taken off line by a fiber cut, it's far more likely to be somewhere between the major backbone and the ISP (a circuit which from the ISP's perspective may be controlled by the major backbone), than it is that the fiber cut will actually isolate the major backbone's POP. The major backbones at this point have a fair amount of redundancy built in, while the circuit from the major backbone to the ISP is likely to be a single circuit on a single path.
Still, even in that environment, most circuit outages are not fiber cuts.
There are no reliable public statistics concerning outage causes for IP networks. The FCC and NRIC have a focus group establishing a voluntary outage reporting process for Cable, IP and Wireless providers. See http://www.nric.org/ Last quarter 49% of all FCC reported outages were facility failures (cable cuts and similar outside plant problems). Other sources of outages were Signalling (21%), CO Power (12%), Local switch (12%) and Tandem switch (6%).