...

John Scudder then discussed some modeling he and Sue Hares have done on the projected load at the NAPs. The basic conclusions are that the FDDI technology (at Sprint) will be saturated sometime next year...

MarkFedor/ColeLibby from PSI said there was a "quiet" admission that the old methodology was already "approved" for the SPRINT NAP.

This technology works *now* (Sprint NAP is already in production use) and there is a clear path for upgrading it to practically unlimited aggregate bandwidths: 1) replace FDDI concentrators with FDDI switch (this gives us up to 1Gbps of total bandwidth while limiting access to one full DS-3) 2) replace single FDDI with multiple FDDIs (realistically with 2 or 3) -- allowing access at OC-3 speeds 3) build point-to-point bypasses between peers with OC-3s or FDDI to take the traffic out of shared medium. We will hit limitations of the present/then-future routing technology long before we'll exhaust the possibilities to increase the NAP aggregate and access bandwidth by cheap incremental upgrades. That's the whole point of Sprint NAP architecture. ATM/SMDS/Flame Delay do not get even close to what we achieve (and ATM is not useful yet). Load balancing between NAPs is necessary, but for entirely different reasons -- first, because of capacity limitation on nation-wide backbones and, second, to reduce latencies on cross-ISP traffic. --vadim