1.

A.

Same as [[2017_03]]

B.

Same as [[2017_03]]

C. (6p)

Fill in the prefixes belonging to the descriptions:

• 6to4: 2002::/16
• Global unicast: 2000::/3
• Link-local unicast: fe80::/10
• Localhost: ::1/128
• Multicast: ff00::/8
• Unique local unicast: fc00:/7

D.

I.

Not relevant

II. (4p)

What is the IPv6 network with the largest prefix length that includes all of 2017:1219:8800::/37, 2017:1219:a000::/36 and 2017:1219:bc00::/38

• 2017:1219:8800::/37 → third hextet = 0x8800 (binary of top nibble 8 = 1000)
• 2017:1219:a000::/36 → third hextet = 0xA000 (top nibble A = 1010)
• 2017:1219:bc00::/38 → third hextet = 0xBC00 (top nibble B = 1011)

The first two parts are identical, so at least a /32 prefix
The first 2 bits of the third hextet are also identical, so a /32 prefix
That means the last hextet has to start with 1000 -> 8
So 2017:1919:8000::/34

2.

A.

I. (2p)

SNAP was not covered

II.

SNAP was not covered

III. (2p)

What is a VLAN and in what specification is it defined?

VLAN stands for Virtual Local Area Network, allowing the segregation of hosts and traffic within a single physical LAN

It is defined in 802.1q

IV.

Not relevant

B.

Not relevant

C.

D. (5p)

• Top right has no root port, meaning it is the root bridge, thus ID 1
• Bridge 2 has equal cost paths through all other bridges, but prefers bottom right, thus that one has the lowest ID of the three, which would be 3
• The network between the center and top left node prefers the top left node, indicating it has a lower ID, which means top left is 4 and center is 5

Thus:

• Top left: 4
• Top right: 1
• Bottom right: 3
• Bottom left: 2
• Center: 5

3.

A.

Not relevant

B.

I. (2p)

Which two OSPF LA types distribute topological information?
Which routers originate this information?

The two types are Router LSAs, generated by all OSPF routers and network LSAs, generated by the designated router

II. (2p)

Which two OSPF LA types distribute prefix information?
Which routers originate this information?

The two types are summary LSAs, generated by the area border router and external LSAs, generated by the AS border router

C. (5p)

Draw an OSPF network in which you show a typical stubby area, a typically totally stubby area and a not so stubby area

                  [ Backbone Area 0 ]
                   /               \
      (stub) Area 1                 Area 2 (totally stubby)
                   \
                    Area 3 (NSSA)

Area 0: The backbone (normal area).
Area 1 (Stub):
No Type 5 LSAs (external routes) allowed in.
ABR injects a default route (Type 3 LSA).
Area 2 (Totally Stubby):
No Type 3, 4, or 5 LSAs from other areas.
Only default route from ABR.
Area 3 (NSSA):
Allows local external routes (Type 7 LSAs generated by ASBR inside).
ABR translates Type 7 → Type 5 for other areas.

D.

I. (4p)

Draw the result of Dijkstra

Loop graph, directional arrows clockwise, nodes a-h counter-clockwise

The algorithm can will update direct neighbours each step, all other distances are considered infinite, thus it must follow the direction of the arrows:
a -> h -> g -> f -> e -> d -> c -> b

II. (2p)

Suppose you are allowed to make ONE edge bidirectional, for which edge will the result be different

Any edge in the middle of the loop will not change the graph, since going back over the edge we've already traversed will not lead to a shorter distance

Therefore, the only edge which will change the outcome is the edge of a->b, since we can then directly traverse that from the source, meaning we actually get a shorter path to B (cost 1 instead of cost 35 when going around)

4.

A. (3p)

Which of the following BGP attributes are "well-known mandatory"?
(Multiple answers possible)

• AS_PATH
• COMMUNITY
• MED
• NEXT_HOP
• ORIGIN

AS_PATH, NEXT_HOP and ORIGIN (that always has a value of 0 nowadays)

B. (5p)

Complete the following list for BGP route selection preference:

1. Weight
2. ...
3. ...
4. Lowest Origin
5. ...
6. ...
7. Lowest IGP cost to BGP egress
8. ...

1. Weight
2. Highest LOCAL_PREF
3. Shortest AS_PATH
4. Lowest Origin
5. Lowest MED
6. eBGP over iBGP
7. Lowest IGP cost to BGP egress
8. Oldest route

C.

Same as in previous exams

D.

Why is iBGP scaling needed? (3p)
Name two methods (2p)
Suppose we have 10 speakers:
How many sessions do we need for a full mesh (1p)
What is the minimum amount for method 1 of iBGP scaling? (1p)
What is the minimum amount for method 2 of iBGP scaling? (1p)

Same as in previous exams: full internal mesh does not scale
Can be fixed with route reflectors, or fake internal ASes that speak BGP
For 10 speakers we need $10\ast 9/2=45$ sessions
With route reflectors, every speakers still needs an iBGP session to the reflector, so 9
With internal ASes, every speaker can be it's own fake AS, so we can use 0 iBGP