1. How Will the Optical Internet Evolve? Joseph Bannister Computer Networks Division USC Information Sciences Institute SPIE Photonics East Boston, Mass. September 20, 1999

2. 29/18/99University of Southern California Information Sciences Institute Panelist’s Disclosure =IP-centric researcher with earlier incarnation as optical-networking researcher =Also runs an ISP (Los Nettos) =Familiar with Internet operations and infrastructure deployment

3. 39/18/99University of Southern California Information Sciences Institute IP Over ATM =Widely deployed in Internet backbones for data communications =Ease of management =Permanent virtual circuits and paths =High overhead and fragmentation issues for IP =Quality-of-service model largely ignored in practice

4. 49/18/99University of Southern California Information Sciences Institute IP Over SONET =Lowered overhead (with respect to IP Over ATM) =Not as flexible for operators as IP Over ATM =Cost of add/drop multiplexors is high

5. 59/18/99University of Southern California Information Sciences Institute IP Over WDM =Near-term outlook =Link multiplier =Medium-term outlook =Virtual WANs =Backbone peering =Long-term outlook =Label switching =Megarouters

6. 69/18/99University of Southern California Information Sciences Institute WDM in the Near Term =Squeeze more out of a long-distance fiber =Simple multiplexing of data on an expensive optical-fiber link

7. 79/18/99University of Southern California Information Sciences Institute WDM in the Medium Term =Operate several SONET rings on a single physical mesh =WDM’s role in peering International Backbone National Backbone National Backbone National Backbone WDM Internet Exchange Point Peering Matrix

8. 89/18/99University of Southern California Information Sciences Institute WDM in the Long Term =Optical label switching =Lack of success with ATM label switching =Wavelength shortage =Optimized virtual topologies =Centralized schemes: that dog won’t hunt =Rise of megarouters =Physical issues (dispersion, amplification, propagation delay)

9. 99/18/99University of Southern California Information Sciences Institute Switching Performance

10. 109/18/99University of Southern California Information Sciences Institute Switching Performance with Merging

11. 119/18/99University of Southern California Information Sciences Institute Routing in a Rich Topology =Wavelengths can be used to create topologies with high degrees of connectivity =ATM-based backbones use this today =More than 10 routers with degree greater than 1000 were discovered (“megarouters”) =Effects of megarouters on routing protocols are not completely understood

12. 129/18/99University of Southern California Information Sciences Institute Megarouters Sparse Physical Topology Highly Connected Virtual Topology