Background reading from Fiberoptics Online:
Petabit-Class Capacity to Make Moot Router/Cross-Connect Debate
10/26/99 By: Erik Kreifeldt
Although not the most fervent debate that the conference, the terabit router/wavelength cross-connect debate continued at the 22nd Annual Newport Conference on Fiberoptics Markets (Oct. 18-20; see Conference Chronicles New Trends in Fiber Deployments and Fiber Leasing Becomes More Attractive).
"The technologies are not mutually exclusive?they're both essential," offers Williams Network CTO Matthew Bross. With tens of thousands of wavelengths to manage in a national network, optical switching will be required. On the edge, electrical aggregation of bits will require a terabit router, he says, adding that terabit-class capacity will more likely exist at the edge of the network than the core?where he predicts petabit-class capacity will ultimately reside.
Terabit router proponents, such as Pete Chadwick, director of marketing for Avici Systems, reason that as the majority of communications traffic becomes Internet protocol (IP), it makes sense to optimize equipment to carry IP and map legacy protocols onto the IP traffic. Chadwick offers a prediction that 95% of network traffic will be IP by the early '00s.
But legacy carriers and even new carriers are not pure IP plays, observes Shannon Hansen, senior product marketing manager for Cisco Systems' Monterey Networks. A wavelength router accommodates optical signals from backbone IP routers, asynchronous transfer mode (ATM) switches, and SONET add/drop multiplexers, and manages these signals at the optical layer independent of their protocol.
Terabit router proponents must map non-IP services onto an IP platform, a stunt that bears an unreliable perception compared with circuit-oriented approaches, Chadwick acknowledges. Nevertheless, he maintains that cross-connects consume more bandwidth and generate more complexity than terabit routers, because cross-connect networks support many gigabit-class routers in lieu of few terabit routers, and comprise another network layer, albeit one optimized for transport.
Configuring a national network of wavelength cross connects in a mesh architecture saves carriers about half the cost of a national network of SONET rings, Hansen calculates, maintaining that these economics prove in the approach. A combination of routers with wavelength cross-connect equipment is only more cost-effective than terabit router deployments when the port cost of the routers is more than 1.5 times the port cost of the cross-connect, Chadwick calculates
|
