BOOM! there goes WorldCom

Sprint uses 4 ring sonet loops. Two backup and two active. This way
there is total capacity availiable in the event of a fiber cut. The data
just takes the reverse way around the loop on the backup path. I believe
they should have all the loops completed by the end of the year.

Most fibre point-to-point links use two fiber strands, and typically
one is set up as a transmitter and the other as a receiver.
That is, two unidirectional fibre strands form a pair.

(With WDM, one ends up with multiple wavelengths on the same fibre --
which is roughly akin to cleaving the fibre into two fibres -- and one
can either have multiple transmitters or mixed transmitters and
receivers on any single strand. A single strand, therefore, can
look like a fibre pair, if you wanted, or a pair of strands could
carry multiple "virtual fibre strands" each containing light at
different frequencies)

A fully protected SDH or SONET ring is made up of two or more
pairs of fibre along different physical paths. Each SDH or SONET
multiplexer has sufficient buffering to make up for any propagation
delay difference (sometimes many milliseconds thanks to the slowness
of light in glass, and also thanks to repeater jitter). A copy
of the data recently arrived on the shorter physical path is retained
in memory and compared against the copy which arrives later along the
longer path. Any discrepancies are flagged as errors, and set off
alarm bells. Missing data is considered a line failure, and if data
is missing off the primary path, the data arriving from the secondary
path is forwarded instead.

SDH and SONET therefore are theoretically immune to any single
equipment failure outside of the MUXes themselves. The time to
recover from any fibre cut or optics failure is roughly the difference
in the observed propagation delay plus a fudge factor which takes
thermal and other effects into account.

With the ongoing deployment of WDM in carriers' networks, there
will be moves either to parallelize SONET and SDH using the
traditional four strands, to do clever aggregation (present a
concatenated unidirectional fibre to a MUX at the cost of some
further propagation delay and antijitter techniques), or to deploy
2-stranded SDH/SONET rings. There will also be pushes to use
multiplexing technologies other than SDH or SONET.

Which is the best approach depends significantly on point-of-view.
Personally I like the idea of ever-faster bit- (or byte-) pipes
in an effort to keep the number of interfaces in switching equipment
down. Others strongly disagree and would make the opposite trade-off.