This problem examines effects of SONET protection switching on applications.
Assume a 3-node SONET ring among Minneapolis, Chicago, Denver.
Relevant distances: MSP->Chi = 570km, Chi-Den = 1466km,
Den-MSP = 1115km.
Assume UofMN has acquired OC-48 (2,488mbps) bandwidth on that SONET ring, between Minneapolis and Chicago.
Working bandwidth is MSP<-->Chi;
Protect bandwidth is MSP<->Den<->Chi
For hosts to "fill the pipe" for a TCP connection,means that they must have TCP send/receive buffers at least as large as the one-way bandwidth*delay product of the path traversed by the TCP connection.
Assume host1 in MSP, host2 in Chi each have send/recv buffers of 2MBytes (this is fairly large; linux kernels default to 64k).
a. What is the bandwidth-delay product for the normal, "working"path MSP<-->Chi (1-way, in bytes)?
b. What is the bandwidth-delay product for the "protect"path MSP<-->Den<-->Chi (1-way, in bytes)?
c. Assume the hosts are connected at OC-48 (2,488mbps),and are transferring a large file MSP->Chi.
Under normal operating conditions, should the hosts "fill the pipe"?
Why/why_not?
d. If the MSP-Chi link is destroyed, and the SONET ring "wraps",will the hosts "fill the pipe"? Why/why_not?
e. Explain what happens in "d"; describe (roughly) what you'd expect to see on the link leaving host1.