Solutions to Homework #2
Chapter 3: 21, 22, 23
21. (10 pts)
An Ethernet frame has a minimum frame size (5 pts) of 64 bytes; it
pads smaller packets. Protocols above the Ethernet must not confuse
padding with actual data, so it must remember where the data ends (5 pts).
22. (10 pts)
The calculated delay is 1500/2*10^8 = 7.5us (2 pts). The 51.2us value
is conservative, and reflects several factors.
- Original Ethernet limited to 1500m, now up to 2500m. (1 pt)
- 51.2us reflects round trip time, not one-way latency. (5 pts)
- Takes latency introduced by repeaters into account. (2 pt)
23. (10 pts)
A remote terminal is a latency-dominated application (4 pts); a small
amount of data is sent in each packet. The Ethernet is better because,
except under heavy load, the time it takes to acquire the network is small
(ie no waiting for token). FDDI is better suited for file transfer because
such an application is higher bandwidth and 100MBps is better than
10Mbps (4 pts). Whether the same holds on a 100Mbps ethernet depends on
the network load. Under high load, FDDI is more efficient for file
transfer (2 pts).
Chapter 4: 1, 2
1. (10 pts)
Datagram-happy applications are ones that can handle packet loss and
retransmission, like FTP, SMTP, HTTP, etc. (5 pts)
Connection-requiring applications are ones like real-time applications
that the data MUST get there on time (real time voice, video, etc.) (5
pts)
2. (10 pts)
Subpart | Switch | Input Port |
Input VCI | Output Port | Output VCI |
a |
| 1 | 2 | 0 | 1 | 0 |
| 2 | 3 | 0 | 0 | 0 |
| 3 | 0 | 0 | 3 | 0 |
b |
| 1 | 3 | 0 | 1 | 1 |
| 2 | 3 | 1 | 1 | 1 |
| 4 | 1 | 1 | 1 | 0 |
c |
| 2 | 2 | 0 | 0 | 1 |
| 3 | 0 | 1 | 2 | 0 |
d |
| 1 | 0 | 0 | 1 | 2 |
| 2 | 3 | 2 | 0 | 2 |
| 3 | 0 | 2 | 3 | 1 |
e |
| 4 | 2 | 0 | 3 | 0 |
| 2 | 1 | 0 | 0 | 3 |
| 3 | 0 | 3 | 1 | 0 |
f |
| 4 | 0 | 0 | 3 | 1 |
| 2 | 1 | 1 | 3 | 0 |
| 1 | 1 | 0 | 2 | 0 |
Chapter 5: 15
15. (10 pts) Since the lower level network is a connection oriented one,
there
would be a connection set-up and tear-down phase. This would imply that IP
would need to set-up and tear-down a connection for each packet it
forwards. This would be very inefficient (8 pts). To solve this problem,
IP might cache (and reuse) the recently used connections, and close down
the least recently used connection whenever it needs to open a new
connection (2 pts).
In addition:
1. (5 pts) Baseline drift if there are long idle periods (3 pts), but NRZI
encoding is used (2 pts) so no problem.
2. (10 pts) A CRC's error correcting capability is based on its ability to
detect more than one error; the reason frames are simply
discarded is because recovery is very computationally intensive
and burst errors are hard to differentiate from correctable
single-bit errors.
3. (5 pts) Easier to implement in hardware.
Bonus Question
(5 pts) If two stations transmit claim frames with TTRT equal to the
lowest TTRT on the ring, they'll both get them back and assume that they
hold the token.