CS/ECE 438: Communication Networks Fall 2007

Midterm Exam Solutions

1. (34 points) Short Answer

(a) (2 points) How many IP addresses are there in a /27 network?

25 − 2 = 30

two addresses for broadcast and network name.

(b) (2 points) Which is more efficient, a Manchester encoding or a 4B/5B encoding?
Manchester encoding is only 50% efficient whereas 4B/5B is 80% efficient.
(c) (2 points) Why does the 802.11 wireless Ethernet protocol not use CSMA/CD?
In wireless networks, it is impossible to detect collisions during transmissions because
the receiver radio has to be turned off for the duration.
(d) (3 points) What is the hidden terminal problem?
If B can hear both A and C, but A and C cannot hear each other due to being out of
range, A may start an interfering transmission to one already in progress from C to B,
creating a collision at B that A cannot detect.
Note: A may be transmitting to another node D, rather than C, and a collision would
still occur at B.
(e) (3 points) How does 802.11 solve it?
The transmitting node sends a RTS and the receiving node sends a CTS before transmis-
sion, indicating the period that the channel will be busy. Nodes that can hear the receiver
will hear the CTS and stay silent for that period of time.
(f) (12 points) List one advantage and one disadvantage each for the following forwarding
schemes:
i. Datagram forwarding
Advantages: Ability to re-route packets dynamically in case of failures.
Can send traffic immediately.
Disadvantages: Header requires full unique address.
May not be possible to deliver packet
Successive packets may not follow the same route
Global address to path translations require significant storage.
ii. Virtual Circuits
Advantages: Header requires only a small VCID.
Can reserve resources when setting up connection.

Disadvantages: Need to wait 1 RTT per set up.

Failure require the connection to be re-established.
iii. Source Routing
Advantages: Simple switches.
Route under control of the source.

Disadvantages: Hosts must know entire topology (and keep it up to date).

Headers might become large.
NETID: CS/ECE 438 Midterm 2

(g) (6 points) List one advantage each for: Link State Routing, Distance Vector Routing.
Link state routing: Global view of the network allows for more optimizations, multiple
path selection.
More robust to errors.
No count-to-infinity problem.
Distance-vector routing: Potentially smaller message traffic.
Less state maintained at nodes.
An updated link cost does not need to be broadcast to the entire network.
(h) (4 points) List two reasons for using a different routing schemes within an AS (intra-AS)
and between AS’es (inter-AS).
Policy: different goals addressed by routing within the AS (e.g. performance) and between
AS’s (e.g. revenue maximization).
Performance: Can optimize traffic within the AS better because of more limited view and
single administrative domains.
Scale: AS abstraction reduces table size and update traffic, as routing changes within an
AS need not be propagated to other AS’s.
Compatibility / innovation: Intra-AS schemes give AS’s flexibility to choose different
designs and experiment with new ones, whereas inter-AS scheme must be the same for
all routers in the world.

2. (6 points) The USB protocol uses a CRC with the polynomial x5 + x2 + 1. Calculate the
CRC of the bit string 1100011110110001
The polynomial x5 + x2 + 1 is equivalent to the bit string 100101.

100101 | 110001111011000100000
100101
------
101001
100101
------
110011
100101
------
101100
100101
------
100111
100101
------
100001
100101
------
100000
100101
------
10100
NETID: CS/ECE 438 Midterm 3

3. (6 points) Consider 4 hosts on a single LAN with the following ARP tables:

10.0.0.1 10.0.0.2 10.0.0.3 10.0.0.4

MAC: 0:0:1 MAC: 0:0:2 MAC: 0:0:3 MAC: 0:0:4
IP MAC T IP MAC T IP MAC T IP MAC T
10.0.0.3 0:0:3 5 10.0.0.4 0:0:4 7 10.0.0.1 0:0:1 5 10.0.0.2 0:0:2 7

The third column, T , represents the number of minutes since the ARP cache entry has
been updated. Suppose 10.0.0.1 performs an ARP request for 10.0.0.4. After the request is
complete, its table is:
10.0.0.1
MAC: 0:0:1
IP MAC T
10.0.0.3 0:0:3 6
10.0.0.4 0:0:4 0
Write the contents of the ARP tables for the other three hosts.
Host 10.0.0.4 will add an entry for 10.0.0.1 to its ARP cache, since it will likely need to
communicate with 10.0.0.1.
Host 10.0.0.3 will snoop on the ARP request and update the T field for its entry for 10.0.0.1
to 0.
Host 10.0.0.2 will also hear the ARP request, but because it has no entry for 10.0.0.1, it
will ignore the request. 10.0.0.2 will not hear the ARP response, because ARP responses are
unicast. So the tables will be:
10.0.0.4
10.0.0.2 10.0.0.3
MAC: 0:0:4
MAC: 0:0:2 MAC: 0:0:3
IP MAC T
IP MAC T IP MAC T
10.0.0.2 0:0:2 8
10.0.0.4 0:0:4 8 10.0.0.1 0:0:1 0
10.0.0.1 0:0:1 0

4. (7 points) Consider the following Banyan network:

NETID: CS/ECE 438 Midterm 4

101

110

(a) Draw the path taken by the packet 101 (shown) towards the output port
(b) Show (on the diagram) how another packet on another input, destined for a different
output port, can collide with 101 inside the Banyan network
(c) What condition would prevent such a collision?

For (b), note that for a collision, it is necessary that the two packets occupy the same wire,
rather than simply arrive at the same 2x2 switch. So if the packet 110 was instead 010, no
collision would have happened.
For (c), sorting the input packets in ascending order can be shown to prevent collisions. A
Batcher network can perform the sorting.

5. (10 points) Consider a client connecting to a server, sending a request, and then reading
back a response. Show the sequence of networking system calls that must occur on the client
and the server for this to work, in the order that they must be executed. The calls you should
use are: accept, bind, close, connect, listen, read, socket, write. You do not need to
list the arguments to the calls.
Client Server
socket socket
bind
listen
connect accept
write read
read write
close close
Note: We didn’t grade the ordering of calls between the client and the server because it is ill-
defined. For example, accept may be called either before or after the connect call; whichever
NETID: CS/ECE 438 Midterm 5

is called first will simply block until the other one is called. (But if you call connect before
listen is called, you will get a “connection refused” error.)

6. (12 points) Consider hosts A and C connected by two Ethernet segments, with a bridge B in
the middle. The segments are 10 Mbps and introduce a latency of 3 µs and 7 µs respectively.

Bridge

A B C

7 μs 3 μs

A is transferring data to C by sending Ethernet frames with a 1500-byte payload. After

sending a frame, it waits for an acknowledgment from C before sending the next one. The
acknowledgment has a 10-byte payload.

(a) Recalling that an Ethernet frame includes an 8-byte preamble and a 14-byte header,
calculate the effective bandwidth of the data transfer.
Recall that a bridge is a pass-through device that introduces almost no processing delay.
So in this case, when A sends a bit, it arrives at C 10µs later. The time to send a whole
packet is then:

(1500 + 14 + 8)bytes
+ 10µs = 1227.6µs
10M bps

The time to send the acknowledgment is:

(10 + 14 + 8)bytes
+ 10µs = 35.6µs
10M bps

Therefore, the effective bandwidth is:

1500bytes
≈ 9.5M bps
1227.6 + 35.6µs

(b) Calculate the effective bandwidth if B was replaced by a switch.

A switch will buffer a packet before sending it out. So the total transfer time of a packet
from A to C is: transmission time at A + latency A to B + transmission time at B +
latency B to C. To send the forward packet, it takes:

(1500 + 14 + 8)bytes
2 + 10µs = 2445.2µs
10M bps
NETID: CS/ECE 438 Midterm 6

To send the ack, it takes:

(10 + 14 + 8)bytes
2 + 10µs = 61.2µs
10M bps

The effective bandwidth is:

1500bytes
≈ 4.8M bps
2445.2 + 61.2µs

(c) Calculate the effective bandwidth if the switch was cut-through.

A cut-through switch acts nearly like a bridge in that it does not buffer packets. However,
it needs to receive the full header before deciding which port to output on. Therefore, it
22bytes
adds a processing delay of 10M bps = 17.6µs in each direction. The effective bandwidth is
then:
1500bytes
≈ 9.2M bps
1227.6 + 35.6 + 17.6 ∗ 2µs