Problem #1

(a)

round trip delay = 16ms (@8 microsec/mile)

For peak TCP rate:

CR=16 * 10 ^–3 * 10 * 10 ⁶ / (48 * 8) = 416

Note: AAL5 payload = 48 bytes/cell

(b)

buffer must be 416 * 53 = 22Kbytes

(c )

Problems:

Buffer size on node B (100VCs would need 2.2Mbytes)

Control overhead

• In a knockout switch configuration, the AAL5 traffic may need to be re-routed to a separate knockout switch entirely, in order to avoid interference

Problem #2

(a)

Bellman Ford can be used to propagate minimum residual bandwidth info in the same way and along with min path.

(b)

Each source simply sets its rate to a factor a (e.g. ½) of the advertized residual bandwidth

Problem #3

Pros of multifiber:

• Faster, cheaper than tunable filters/receivers
• Reduction in path attenuation by properly configuring fiber subtrees
• Each fiber can be further subdivided via WDMA (S/T/WDMA)
• Better reliability by diverse routing

Cons of multifiber:

• Less flexible (multi/broad cast subtree constraints)
• Arduous physical reconfiguration
• Additional cost of switched at internal nodes
• Additional cost of switch control network
• Additional fiber cost (albeit small marginally)