The explosion of data traffic and the availability of enormous bandwidth vi
a dense wavelength division multiplexing (DWDM) and optical amplifier (OA)
technologies make it important to study optical layer networking and restor
ation. This paper is concerned with fast distributed restoration and provis
ioning for generic mesh-based optical networks. We consider two problems of
practical importance: determining the best restoration route for each wave
length demand, given the network topology and the capacities and primary ro
utes of all demands, and determining primary and restoration routes for eac
h wavelength demand to minimize network capacity and cost The approach we p
ropose for both problems is based on precomputing. For each problem, we des
cribe specific algorithms used for computing routes. We also describe endpo
int-based failure defection, message flows, and cross;connect actions for e
xecution of fast restorations. Finally we report test results for large car
rier-scale networks that include both the computational performance of the
optimization algorithms and the restoration speed obtained by simulation. O
ur results indicate that subsecond restoration, high capacity efficiency, a
nd scalability can be achieved without fault isolation and with moderate pr
ocessing. We also discuss methods for scaling algorithms to problems with v
ery large numbers of demands. The wavelength routing and restoration algori
thms the failure detection, and the message exchange and activation archite
ctures we propose are collectively known as WaveStar(TM) advanced routing p
latform.