1.

A.

I. (4p)

What are the two different purposes IPv4 addresses are used for in the current internet architecture?

IPv4 addresses are used as both a location identifier and an endpoint identifier

Location: knowing where to go, mostly between networks
Endpoint identifier: knowing who/what a host is

II.

Which of the following protocol try to separate the two purposes?
(Multiple choice, multiple possible)

• BGP
• HIP
• ILNP
• IPv6
• LISP
• OSPF
• STP

Host Identity Protocol (HIP), Identifier-Locator Network Protocol (ILNP) and Locator/ID Separation Protocol (LISP)

B.

I. (5p)

Consider three LANs, two hosts and two routers ... draw a bipartite graph

LANs on one side, all other devices on the other

II. (3p)

Suppose LAN 1 is in fact a switch S, which also has an IP address, how would you represent S in your picture

Present S as a device AND a network, where the device will be with the other devices, all devices connected to L1 will be connected to it's "network"

C.

What the hell

2.

A. (6p)

Describe the different kinds of equipment that are being used on the layers 1, 2 and 3 in a network environment and describe their characteristic properties relating to collisions and broadcasts

L1:

• Devices: cable, dumb repeaters
• Collisions: possible in star-shaped networks, not a problem with full duplex peer-to-peer
• Broadcasts: repeaters can enable broadcasting on L1, but with direct peer-to-peer links there is no difference
  L2:
• Devices: NICs, switches
• Collisions: not a problem due to separate collision domains after the introduction of switches instead of dumb hubs/repeaters
• Broadcast: possible by sending to MAC FF:FF:FF:FF:FF:FF, but can also unicast to a specific host
  L3:
• Devices: routers, end-user devices
• Collisions: handled by lower layers
• Broadcasting can be done to IPv4 255.255.255.255, unicast is also possible. Routers form a boundary to broadcasts, creating separate broadcast domains

B.

Not relevant

C.

1. Root bridge: bridge with lowest bridge ID
2. Root ports: shortest path, if equal determine based on bridge ID of peer
3. Designated port: same idea

D.

Given the following result of an STP run, what can you deduce about the bridge IDs?
2 is give, place 1, 3, 4 and 5

• We can see there is only 1 bridge without a root port (top left), this must be bridge ID 1
• 2 chose it's root port to go through bottom left, therefore bottom left has a lower ID than top right and middle, thus bottom left is 3
• Both the networks on the right chose top right as their designated port, thus top right must have a lower ID than center. Therefore top right is 4 and center is 5

So:

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

3.

A.

I. (2p)

What information is passed on in a RIP version 1 packet?

RIP packets include the distance (hop count) to each destination currently known to the source
Since there is no next hop field, the advertiser is the next hop

II. (3p)

Why isn't the gateway part of this information?
How is the gateway information obtained in RIP version 1 and 2?

The gateway is not in this information since the advertiser of the route is considered the gateway in RIP v1
In RIP v2 a next-hop field was introduces, enabling advertiser to bypass themselves and point traffic directly to the gateway

III. (2p)

How does RIP v2 use multicast? And RIPng?

RIPv2 multicasts to 224.0.0.9 so that only RIP speakers receive these packets
RIPng is meant for IPv6 and uses the FF02::9 prefix to multicast

IV. (1p)

What information in RIPv2 is disguised as a fake route update?

Authentication can be sent as the first route entry, since there is no separate (header) field for it
It can be identified with address family 0xFFFF

B.

I. (3p)

Name three shortcomings of RIP that were addresses by IGRP

1. Hop-count as a metric is not always accurate (e.g. one slow link vs multiple fast links) IGRP uses a composite metric that is more accurate
2. RIP had a size limit of 15 hops, IGRP has increased this to 255
3. RIP is less scalable, since all peers talk to one another, IGRP can do hierarchical routing by splitting into sections (called ASes, not the same as in BGP)

II. (3p)

Name three additions EIGRP added to IGRP

1. Faster convergence (and loop-free) by using DUAL
2. No periodic updates, only triggered updates -> less noise on the network
3. Introduction of ACKs for reliable delivery

C.

Not relevant

D.

Do Dijkstra

4.

A. (5p)

What are the tree most important BGP attributes and selection criteria used in path selection in order of preference

1. Highest local preference
2. Shortest AS_PATH length
3. Lowest MED

B.

I. (3p)

What does hot potato routing mean and why is this relevant?

Hot potato routing refers to the practice of getting packets out of your network in the shortest IGP distance possible

It is also referred to as "selfish routing", since you force others to compensate for the fact you want the packet out of your network

It can also have unintended consequences by forcing traffic to go over a slow link, thus being slower than using normal or cold potato routing

II. (2p)

Which classical attribute is used to prevent your upstream from using hot potato routing? Is this always successful?

The attribute in question is the Multi Exit Discriminator. It can be used to indicate a preferred link, in case there are multiple present.

It will work as long as your peer respects it, but there is no guarantee, since it is up to their own policies which route is used in the end

III. (2p)

What attribute was introduced later to communicate preferences to your upstream or to receive extra routing information from your upstream?

The new attribute is called communities

It is based upon agreements between parties, that define what certain values mean
It can be used to (for example) request a certain value for local preference, or to not advertise certain routes further

C.

I. (2p)

Why is iBGP scaling needed?

Since routes learned through iBGP do not get forwarded to other iBGP peers, a full internal mesh is needed to ensure routes propagate to all internal peers.

This quickly becomes unmaintainable, and therefore scaling techniques such as route reflectors are needed, as they remove the need for a full internal mesh

A route reflector breaks the split horizon by actually re-advertising the routes to other iBGP peers. Peers still do not advertise to one another

II. (4p)

What methods are available to implement iBGP scaling and how do they work?

1. Route reflectors, as explained above
2. Split your AS into multiple "fake internal ASes". These internal ASes will speak eBGP with one another, removing the need for a full internal mesh across ALL nodes (still need one in each internal AS)

D. (7p)

Describe what kind of ISP SURFnet is and how they use BGP.

SURFnet is an ISP that connects educational networks.
They function as an internet exchange point and will therefore use BGP in the following ways:

• As any normal ISP would, to advertise their and their customer's routes to other peers
• To function as a transit between connected customers
• Perhaps to also run an eBGP route server, to make it easier for their customers to peer with all other customers/peers/providers