CS/ECE 438: Communication Networks
Spring 2014
Midterm Examination
Wednesday March 19th, 11:00 to 12:15 PM
Welcome to the midterm examination! Please read everything on this page before you begin.
The midterm contains 8 problems.
The value of each problem is indicated below and at the start of each question. You may skip one of
problems {2,3,4,5}, OR two of {6,7,8}. You must do problem 1.
Show all work and reasoning by writing both work and solution legibly, and box all answers. If the
course staff cannot read a solution, no credit will be given. Please write legibly. If we cannot read your
answer, then it is wrong!
The exam will end promptly at 12:15 p.m. by the clock in the auditorium, or, in the absence or failure
of such a clock, by the proctor's watch. Please cease writing immediately when time is called.
Print your name and ID number neatly in the space provided.
Good luck!
NAME:
NETID:
PROBLEM
POINTS POSSIBLE
1
28
2
16
3
16
4
16
5
16
6
8
7
8
8
8
Total
100
SCORE
1
1. High level questions（28 points）
a. You are the administrator of an Ethernet network for dorms on a university campus. There
was a sale in town on cheap hardware, where some crappy vendor cheaped out and instead of
assigning MAC addresses properly, used the same 40-bit prefix and just cycled through the
lower bits. End result: you have just discovered that a number of newly deployed machines
on your network have duplicate addresses.
i. Perhaps merge these two into “You set this equipment up on the network without
being aware of the problem. Briefly describe what symptoms might show up.
ii. Give a good temporary workaround you would use.
b. Suppose you ran an ISP network. What would be some reasons to disable traceroutes on your
network?
c. Currently hosts in the Internet have multiple "names": the IP address, the MAC address, and
the DNS name. Imagine an "alternate reality" where the designers of the Internet had decided
that each host should have only one name: the DNS name. The MAC address is replaced in
the MAC header with the DNS name. IP routers now do lookups on DNS names, etc. Give at
least one benefit and at least one downside to this design approach.
d. Why are IP addresses assigned as a function of the topology? Are MAC addresses also
assigned as a function of the topology -- if not, why not?
e. Define “hub”, “switch” and “router” – what is the difference between these devices? Under
what circumstances would you prefer a hub to a switch?
f. Do ISPs typically configure export from providers to peers? Why or why not?
g. Why is a consistent export clause important in service-level agreements?
h. What is the BGP decision process? Why do operators wish to exercise
control over this process? How exactly do operators influence the
process to achieve their goals?
2
You could write your answer to Question 1 here if the space of former page is insufficient.
3
2. Error detection（16 points）
a. The USB protocol uses a CRC with the polynomial x^5 + x^2 + 1. Calculate the CRC of the
bit string 10011001. You need to fully show your work to get credits. In other words, credits
won’t be given if only answers are presented.
b. Assume bit string 1011101111000 is received (using the same generator G as in a), determine
if error could be detected or not.
c. Consider a frame consisting of four bits. Assume that the probability of error is 10^(-3),
which is independent for each bit. What is the probability that the frame is received correctly?
d. Continuing with the above four bit case and use the single parity check, calculate the
probability of you determining there are error(s) given the received single parity? Assume the
appended single parity is always received correctly.
a. 10011001 11100
b. The remainder 11111, error is detected
c. (1 - 10^4)^4
4
4
d. ( ) ∗ (10−3 ) ∗ (1 − 10−3 )3 + ( ) ∗ (10−3 )3 ∗ (1 − 10−3 )
1
3
4
3. Distance vector（16 points）
This question deals with distance vector routing on the network shown below. Assume that the
distances shown in the figure do not change over time except as specified in the problem. The nodes
operate in lockstep, in two-phase rounds. In phase one, a node updates its distance vector according to
any changes it heard about in the previous round. In phase two, a node sends out its new distance
vector, if its distance vector changed during phase one.
At the start of round 1, all nodes' distance vectors contain only the links to their neighbors. (They all
send out their distance vectors in phase 2 of round 1.)
The CS department at UIUC bought new Sun Fire V210 servers. They decided to run a distancevector protocol for routing between these servers (even though it is a rather small network). They are
currently configured as the picture below, with respective edge costs.
The CS staff asked for your help. Write down each step of building the distance- vector routing table
for ‘Eve’ so they can compare it to the output of their implementation. You can use abbreviations e.g.,
‘A’ for Adam and ‘E’ for Eve.
The initial routing table at node A is:
a. Show the initial routing table of node E:
Destination
Cost
Next Hop
A
∞
---
B
∞
---
5
C
11
C
D
17
D
b. Show the routing table of node E after one iteration of the algorithm:
Destination
Cost
Next Hop
A
32
D
B
29
C
C
11
C
D
16
C
c. Show the routing table of node E after two iterations of the algorithm:
Destination
Cost
Next Hop
A
31
C
B
28
C
C
11
C
D
16
C
d. In some failure situations, the administrator notices that it takes an exceptionally long time
for the routing protocol to stabilize in this network.
(i) What problem with the distance vector protocol is the cause?
count-to-infinity
(ii) The administrator is told that BGP does not suffer from this problem. What prevents BGP
from having this problem?
path-vector
6
4. Link-state routing（16 points）
This question deals with link state routing on the network shown below. Assume that the distances
shown in the figure do not change over time.
Again, this problem uses rounds as in problem 3. At the start of the first round, nodes only know of
the links to their neighbors. In the first phase of a round, nodes send their neighbors their own Link
State Packet (LSP) and all LSPs they have received. Node x's LSP contains “this LSP is from x, and
x's neighbors are … with costs ...”. In the second phase, the nodes calculate shortest paths based on
their current knowledge.
a. List the LSP’s (by host name) held by host A at the beginning of rounds 1,2,3,4.
b. Note that although the measured distances in the figure are the same in both directions, the
link state algorithm may NOT assume such to be the case when calculating routes. With this
fact in mind, show host A’s routing table as a list of (destination, next host, distance) 3-tuples
after route recalculation at the beginning of rounds 1,2,3,4.
a. round 1: F, B
round 2: F, B, G, C
round 3: F, B, G, C, D
round 4: F, B, G, C,D,E,H
(also get full points if you write:
round 1: A,
round 2: A F B
round 3: A F B G C
round 4: A F B G C D )
b.
7
Round 1:
B
C
D
∞
E
∞
F
∞
G
H
∞
5
∞
1
Round 2
B
C
D
E
∞
4
∞
F
G
5
3
F
G
4
3
F
G
4
3
F
G
4
3
H
∞
1
Round3
B
C
D
E
∞
4
∞
H
∞
1
Round4
B
C
4
D
E
8
∞
H
∞
1
B
also correct if you go one round further and include:
C
D
E
4
8
9
H
11
1
8
5. Forwarding table and Classless Inter-domain Routing (CIDR) （16 points）
Suppose a router has built up the routing table shown below. The first four lines are for CIDR
addresses, with “/22” indicating a mask of 22 1's followed by 10 0's.
Net/Masklength
128.174.252.0/22
128.174.242.128/25
128.174.252.0/23
128.174.242.0/25
128.174.242.7
Default
NextHop
R1
Interface 1
R2
Interface 0
Interface 2
R3
a. How many individual IP addresses match each of the Net/Masklength pairs, including the IP
addresses for the network address and the broadcast address? For simplicity, ignore the fact that
there is some overlap.
The number of host address possible is determined by the number of bits available for host
addressing (i.e. not used in the network mask). So the formula would be 2x (where x = 32 bits <net mask length>).
Net/Masklength
128.174.252.0/22
128.174.242.128/25
128.174.252.0/23
128.174.242.0/25
128.174.242.7
Hosts
210 = 1024 hosts
27 = 128 hosts
29 = 512 hosts
27 = 128 hosts
20 = 1 host
b. The router can deliver packets directly over interfaces 0, 1 or 2, or it can forward to routers R1,
R2, or R3 according to the table below. Specify the next hop for each of the following destinations.
Remember that if a destination matches more than one line of the table, the longest match is used.
Net/Masklength
128.174.252.0/22
128.174.242.128/25
128.174.252.0/23
128.174.242.0/25
128.174.242.7
default
NextHop
R1
Interface 1
R2
Interface 0
Interface 2
R3
Binary address (net mask in BOLD)
10000000 10101110 11111100 00000000
10000000 10101110 11110010 10000000
10000000 10101110 11111100 00000000
10000000 10101110 11110010 00000000
10000000 10101110 11110010 00000111
i. 128.174.254.252
 Binary address: 10000000 10101110 11111110 11111100
Matches Entry 1: Next Hop R1
ii. 128.174.251.134
9
 Binary address: 10000000 10101110 11111011 10000110
No Match: Next Hop R3
iii. 128.174.242.7
 Binary address: 10000000 10101110 11110010 00000111
Matches Entries 4 and 5: Next Hop Interface 2 (Entry 5 is longer match)
iv. 128.174.242.135
 Binary address: 10000000 10101110 11110010 100000111
Matches Entry 2: Next Hop Interface 1
v. 128.174.255.0
 Binary address: 10000000 10101110 11111111 00000000
Matches Entry 1: Next Hop R1
10
11
6. Channel rate and shared media（8 points）
You are entrusted with the design of a network to interconnect a set of geographically distributed
hosts within your corporation. You have decided to use a copper-based, switched network. The
technology has the following properties: 5MHz signal bandwidth, 511 SNR at transmitter.
What link bandwidth is possible according to Shannon's Law for this network?
C = B*log2(1+S/N)
= 5MHz * log2(512)
= 45Mbps
Using the copper-based network with an 8-point QAM encoding, what modulation rate (baud) is
necessary to obtain the bandwidth found in part a)?
8 point – 3 bits
45Mbps / (3 bits/symbol) = 15 Mbaud
12
7. Interdomain routing（8 points）
Consider the network shown below. Suppose AS3 and AS2 are running OSPF for their intra-AS
routing protocol. Suppose AS1 and AS4 are running RIP for their intra-AS routing protocol. Suppose
eBGP and iBGP are used for the inter-AS routing protocol. Initially suppose there is no physical link
between AS2 and AS4.
a.
b.
c.
d.
Router 3c learns about prefix x from which routing protocol: OSPF, RIP, eBGP, or iBGP?
Router 3a learns about x from which routing protocol?
Router 1c learns about x from which routing protocol?
Router 1d learns about x from which routing protocol?
a. eBGP
b. iBGP
c. eBGP
d. iBGP
13
8. Policy dispute（8 points）
This question deals with policy dispute which was discussed in class. Consider the network shown
below.
Suppose we are concerned with Prefix P in ISP D. The link prices connecting each pair of ISPs are
presented in the above figure as well the local preference of ISP A, ISP B, and ISP C. Describe what
would happen if we run BGP on this topology.
If you need to draw arrows on the graph to illustrate your opinion, please draw them clearly to help us
identify. Just a suggestion, it’s probably helpful for us to understand your answer better if you
perform your analysis based on rounds or stages. But you are definitely not required to do so.
First, ISP D advertise its distance to Prefix P to ISP A, B, and C. ISP A, B, and C learned the direct
connections to ISP D to connect to Prefix P. And each ISP (A, B and C) advertises their distance to
other ISPs.
Second, after receiving new routes, ISP A, B, and C discard the direct connections and chose their
neighbors (e.g. A chose C) in order to connect to ISP D since policies are used to choose routes in
BGP. This leads to the problem that there are no direct routes to ISP D anymore. Next, each ISP
advertises their updated routes.
Eventually, loops appeared in each ISP. For example, A-C-B-A-D-P. BGP kills a route when there is
a loop. So the update will be advertised, and each ISP will finally go back with only choosing direct
connections. And the above processes are going to repeat.
14