Traditional Routing

1
Traditional Routing

• 9th CEENet Workshop on Network Technology
• NATO ANW
• Iskra Djonova Popova (iskra.popova_at_mh.se)

2
Why Traditional Routing?

• This workshop is on mobile and wireless networks
  and you should have a lecture on the routing in
  these networks.
• You need to know where you have been to know
  where you are going.
• To understand routing it is important to know the
  purpose of internetworking.

3
Contents

• Internetworking and the IP Protocol
• Forwarding and Routing
• Routing Protocols
• IGPs vs. EGPs
• Distance Vector Protocols (DV)
• Link State Protocols (LS)
• LS vs. DV

4
Internetworking and the IP Protocol

• Why we need internetworking?
• The goal of an internetwork is
• to build a unified, cooperative interconnection
  of networks that supports a universal
  communication service,
• to detach the notions of communication from the
  details of network technologies, and to hide low
  level details from the user,
• to provide a mechanism that delivers packet from
  their source to their ultimate destination in
  real time.

5
Internetworking and the IP Protocol

• Two main problems are
• Heterogeneity
• The underlying networks can use
• different technologies,
• different media,
• different addressing scheme at the MAC layer,
• different frame formats.
• Scalability
• The Internetwork is not limited in size.

6
Internetworking and the IP Protocol

• IP protocol resolves the problem with
  heterogeneity

H1
H6
App.
App.
Transp.
Transp.
IP protocol
IP protocol
R3
IP protocol
R2
IP protocol
R1
Network
IP
Network
IP
IP
PPP protocol
FDDI protocol
Ethernet protocol
Ethernet protocol
Eth.
PPP
Eth.
PPP
Eth.
Eth.
FDDI
FDDI
Network1(Ethernet)
Network3 (Ethernet)
Network4(point-to-point)
Network2(FDDI)
7
Internetworking and the IP Protocol

• Hierarchical addressing resolvs the problem of
  scalability

Hierarchical addressing
Flat addressing
8
Internetworking and the IP Protocol

• An IP address consists of
• ID of the network to which a host attaches ID
  of a unique host on that network.

• Network address by convention is NetID all 0s

0
31
Network ID
All 0s
9
Internetworking and the IP Protocol

• The subnet mask defines which part of the address
  is the network ID and which is the host ID.
• Consists of contiguous 1s and 0s
• Has the same structure as the IP address.
• 4 bytes written in dotted decimal notation
• Binary 1s in the network and subnet part of the
  address
• Binary 0s in the hosts part of the address

10
Internetworking and the IP Protocol
11
Forwarding and Routing
12
Two Key Routers Functions
Forwarding and Routing
Routing Protocols
Routing (the task of determining the best path
through the network)
Routing Table
Forwarding (the tasks of moving packets from one
interface to another)
Switching
Forwarding Table
13
Forwarding and Routing
Input Ports Processing
Switching fabric
Output Ports
Forwarding Table
Forwarding Decision
Forwarding Table
Forwarding Decision
Forwarding Table
Forwarding Decision
14
Forwarding and Routing

• The routing table consists of at least three
  columns Destnation, Mask (or prefix) and Next
  hop (or interface) column.
• Destination column contains the addresses of all
  possible destinations (networks).
• Next hop column containes the address of the next
  router that must be connected to the same network
  (or the interface to which the the packet should
  be delivered).

15
Forwarding and Routing

• Each IP packet carries its destination address.
• The router performs the following
• Accepts the packet arriving on an incoming link.
• Looks up packet destination address in the
  forwarding table, to identify outgoing port(s).
• Manipulates packet header e.g., decrement TTL,
  updates header checksum.
• Sends packet to the outgoing port(s).
• Buffers packet in the queue and transmits the
  packet onto outgoing link.

16
Forwarding and Routing

• To make the forwarding tables shorter , a special
  network address is used for the default route.
• 0 . 0. 0. 0
• The default route is used when the router doesnt
  have the destination address in its routing
  table.
• The default route reduces the entries of the
  forwarding tables for the routers at the
  periphery of the Internet.
• The forwarding tables at the core are large and
  they are still growing
• They can hardly make use of the default route

17
Forwarding and Routing

• The process of determining the routes in the
  forwarding table is called routing
• The routes can be constructed
• Manually by network administrator.
• Static Routes.
• No dynamic changes to these routes will occur.
• Dynamically by a routing protocol.
• Dynamic Routes.
• Routing information is exchanged between routers.
• The routing metric is used to find the best
  path.

18
Forwarding and Routing
With static routes the router cannot
automatically reroute the packets if a path fails.
B R2 C R3
19
Forwarding and Routing

• Advantages of the static routes
• predictability the administrator knows where
  the packets are going
• no overhead
• simplicity
• Disadvantages
• lack of scalability
• can not adapt to a failure in a network

20
Forwarding and Routing

• When to use a static route?
• When a network is accessible by only one path, a
  static route to the network can be sufficient.
• Configuring static routing to a stub network
  avoids the overhead of dynamic routing.
• The default route is always configured as static
  route.

21
Routing Protocols

• Routers use a common protocol to exchange routing
  information.
• The forwarding table is obtaind using some of the
  routing algorithms.
• Best path between networks is determined by
  Routing Metric.
• Automatic adaptation to topology changes.
• Routing protocols should not be confused with
  routed protocols.

22
Routing Protocols

• Advantages
• Adapt to a failure in a network.
• Work in large networks.
• Disadvantages
• Increase in complexity.
• Overhead on the lines and routers.

23
Routing Protocols

• The goals of the routing protocols is to find the
  best collection of paths for all
  source-destination pairs at every moment.
• The forwarding table is calculated based on the
  routes found.
• Best path is defined in term of a cost
  function.
• Questions asked when defining the best path
• How are the costs defined?
• What are network parameters?
• How are they measured and reported?
• Where are the routes computed and with which
  algorithm?

24
Routing Protocols

• The difficulties when designing a routing
  protocols are
• Some of the parameters that determine the state
  of the network resources are not known at every
  moment.
• Part of the network bandwidth required for user
  services is wasted for distributing the update
  information.
• A compromise has to be achieved (not too much
  information sent, and enough parameters known for
  good routing.

25
Routing Protocols

• The routing protocols design goals should be
  concerned with
• Optimallity
• Simplicity/Low overhead
• Robustness/ Stability
• Rapid Convergence
• Flexibility

26
Routing Protocols

• The routing protocols used today usually
  determine the shortest path tree at each node.
• Different metrics can be used for the lengths
  of the links.
• The information about the network is gathered in
  different ways.

27
Routing Protocols
A network has 6 nodes (A, B, C, D, E, F) and 10
links (a, ..., j). The cost of each link is given
besides the link
Routing table for Node A
E
F
Shortest-path tree with a root at node A
28
Routing Protocols

• The scaling problem
• Over 50 million destinations for the core routers
  even with subnetting and aggregation implemented.
• Storing all destination in forwarding tables is
  impossible.
• Routing table exchange would swamp the bandwidth.

29
IGP vs. EGP

• With the growth of the networks connected to the
  Internet, it become hard to maintain the
  forwarding tables.
• The solution to this problem was split the
  Internet into a set of Autonomous Systems (AS)
• Each AS is a set of routers and networks under
  the same administration
• The routers within the AS use the interior
  gateway protocol (IGP)
• Special routers, called Exterior gateways or
  only gateways are used to connect ASes and they
  use the exterior routing protocols (EGP)

30
IGP vs. EGP
Autonomous System 1
IGP
EGP
IGP
Autonomous System 2
IGP
Autonomous System 3
31
IGP vs. EGP
Host h2
Intra-AS routing within AS B
Intra-AS routing within AS A
32
IGP vs. EGP

• Interior Gateway Protocols (IGP)
• Routing protocol within a single autonomous
  system.
• Single network administration.
• Unique routing policy.
• Make best use of network resources.
• Exterior Gateway Protocols (EGP)
• Routing protocol among different autonomous
  systems.
• Independent administrative entities.
• Communication between independent network
  infrastructures.

33
IGP vs. EGP

• Used inside an Autonomous System.
• Designed to handle more redundant links.
• Links are cheaper in a local environment gt one
  can afford more redundant links.
• Designed with a higher bandwidth in mind
• Cheaper bandwidth gt one can use more bandwidth
  for the exchange of routing information.

34
Distance-Vector Protocols

• The distributed way of learning.
• You need to go to X, but you dont know which
  road to take. You only know the distances to your
  neighbours, A, B and C.
• You get information from your neighbours (A,B,C)
  about their distances to X.

You calculate 42010430 45020470 40017417
minimum (430, 470, 417) 417
35
Distance-Vector Protocols

• Collecting the necessary information
• Each router that uses distance-vector routing
  begins by identifying its own neighbors.
• The port that leads to each directly-connected
  network is shown as having a distance of 0.

• Each of the other network entries in the routing
  table has an accumulated distance vector to show
  how far away that network is in a given direction.

36
Distance-Vector Protocols

• Finding the Least Cost Path
• Nodes send information only to neighbouring nodes
  in regular time intervals.
• The information sent (the distance vectors) are
  all routes (rows) from the routing (forwarding)
  table.
• Routes are installed directly in the routing
  (forwarding) tables, lowest cost wins.
• The higher cost is updated in cases when it is
  advertised by the node that is the next hop for
  that route.

37
Distance-Vector Protocols

• When the topology in a distance-vector protocol
  network changes, routing table updates must
  occur.
• Topology change updates proceed step-by-step from
  router to router.

• The routing tables include information about the
  total path cost and the address of the first
  router on the path to each network contained in
  the table.

38
Distance-Vector Protocols

• Example A network of five nodes (A, B, C, D, E)
  and 6 links (a, b, c, d, e, f). The cost of each
  link is 1. (hop-count is used)
• After initialization each node has a knowledge
  about himself only.
• Each node sends a distance vector to all its
  neighbours.

B
A 0
A
a
b
C
d
A 0
c
e
E
f
D
39
Distance-Vector Protocols

• B adds the cost of his local links and updates
  its routing table
• Other nodes behave in similar way

• After the second update

• D will recieve the distance vector from A and
  update its routing table with an entry for B.
• It will discard the entries for A and D because
  their cost is larger than the costs it already
  has.

A
A 0, B1, D1
B
b
a
d
A 0, B1, D 1
c
C
e
D
f
E
40
Distance-Vector Protocols
The routing tables are complete after third
update
C
A
B
C
b
a
d
c
e
D
f
E
41
Distance-Vector Protocols

• A and B discover this failure, imediately update
  their tabels noting that the link a has an
  infinite cost and all nodes previously reached
  through link a have now infinite distance.

C
A
B
A 0, B inf., D 1 , C inf., E inf.
B 0, A inf., D inf. , C 1, E 1
b

• A and B send their distance vectors to the
  neighboring nodes.

d
C
c
e
D
f
E
42
Distance-Vector Protocols
C
A
B
C
b
d
c
e
f
E
D
43
Distance-Vector Protocols
C
A
B
C
b
d
c
e
f
E
D
44
Distance-Vector Protocols
C
A
B
C
b
d
c
e
f
E
D
The global connectivity has been established
45
Distance-Vector Protocols
A
B

• If link f breaks, the network will be separated
  in two different parts

C
b
d
c
e
f
D

• If D transmits its distance vector to A before A
  transmits its distance vector to D, A will update
  its routing table and notice that all nodes
  except D are unreachable.

E
A
A
A

• If A first has a chance to transmit its last
  distance vector to D, the counting to infinity
  problem will appear. D will update its routing
  table with wrong information

c
D
D
46
Distance-Vector Protocols

• These routing tables have made a loop.
• If during the break of link f, there were packets
  destined to B, C or E in the input queues of D or
  A, these packets will be transfered from A to D
  and back from D to A until their time to live
  expires.
• During this time, link c will be congested, and
  normal traffic between A and D will be annoyed.

A
c
D
47
Distance-Vector Protocols

• A and D will keep exchanging information from
  their routing tables.
• After each exchange the cost will increase.
• This exchange can continue indefinitely unless
  there is a mechanism to stop it.

A

• A cost that corresponds to infinity is usually
  introduced.
• When the costs in the tables reach this cost that
  corresponds to infinity, the looping of the
  packets stops.

c
D
48
Distance-Vector Protocols

• After certain number of iterations the cost to B,
  C and E will reach the value that corresponds to
  infinity.
• A and D now know that B, C and E are unreachable.
• The value for Infinity is cruscial for the time
  neccasary to stop the loop.
• However, the value of infinity limits the size of
  the network

A
c
D
49
Distance-Vector Protocols

• Loops usually occurs when incorrect information
  that contradicts the correct information is sent
  back to a router.
• Logic It is never useful to send info about a
  route back in the direction from which the info
  came!

Split horizon
50
Distance-Vector Protocols

• Two types of split horizon
• Ordinary split horizon ? A router will not send
  an update for a route via an interface from which
  it originally received knowledge of that route.
• Split horizon with poison reverse ? A split
  horizon in which the router responds to attempts
  to update a route with an update that marks the
  route as unreachable (sends un update with
  distance equal to infinity).
• Split horizon resolves only the two node loops

51
Link State Protocols

• Step 1 Each node broadcasts its state to all
  other nodes.
• Step 2 Each node builds a global state
  knowledge.
• Step 3 Each node locally computes the shortest
  paths to all other nodes using Dijkstras
  algorithm.

52
Link State Protocols

• Link
• an interface on the router.
• Link state
• description of the interface and the neighboring
  routers
• IP address, mask, type, routers connected to it.
• Link state database
• collection of link state advertisement for all
  routers and networks.
• Sequence Number
• Shows whether the update sent is a newer one.

53
Link State Protocols

• All routers maintain a complete copy of the
  network map and perform a complete computation of
  the best routes from this local map.
• The network map is held in a database, where each
  record represents one link in the network.
• Each record is inserted by one router that is
  responsible for it.
• It contains an interface identifier, the
  information describing the state of the link, the
  cost, the distance or the metric of the link and
  the sequence number.

54
Link State Protocols
Seq. num.
From To Link Cost A B a
1 2 A D c
1 2 B A a 1
2 B C b 3 1
B E e 2 2 C
B b 3 1 C
E f 1 1 D A
c 1 2 D E e
2 1 E B d 2
2 E C d 2
1 E D e 2 1
Link State Announcement (LSA)
From A to B, Link a, Cost 1, Sequence Num.
2
55
Link State Protocols

• A router generates a new LSA when
• it has a new neighbour,
• the cost of the link to an existing neghbour has
  changed, or
• a link to a neighbour has gone down.
• Every new LSA has a sequence number that is
  higher than the previous one.

56
Link State Protocols

• The purpose of the routing protocol is to adapt
  routes to the changing conditions in the network.
• This can be done only if the link state database
  is updated after each change of the link state.
• When the status of the link changes, it is
  detected by the neighboring routers.
• They change their link state database and
  generate a link state advertisement (LSA) with a
  higher sequence number.
• The flooding protocol is used to transmit the LSA
  to all other routers.

57
Link State Protocols

• The steps of the flooding protocol
• Receive the message.
• If the record is not yet present , add it to the
  database and send LSA on all the links except the
  one it was received on.
• If the record is present in the database and its
  sequence number is lower, replace the record by
  the new value and send it further on.
• If the record is present in the database and its
  sequence number is higher, transmit the database
  value in a new message through the incoming
  interface.
• If both sequence numbers are the same, do nothing.

58
Link State Protocols

• In case of link failures A and B send the
  information to all other nodes about the state of
  link a.
• The flooding procedure distributes this
  information quickly.

• Every router recalculates the shortest path tree
  when there is a change in the link-state
  database.
• The connectivity is reestablished. Loops may
  appear only during the transition period (the
  time during which the link state databases in the
  nodes differ).

59
Link State Protocols
B
C
A

• If b breaks during this period, A and D will not
  receive this LSA and their database will be
  different than the one of B, C and E.

b
3
c
1
2
d
f
1

• When e comes up, A and D has to be synchronized
  with E. This process is called bringing up
  adjacency

E
D

• In the case when network is segmented, the link
  state databases in both parts of the network are
  different.

60
Link State Protocols

• Exchanging complete copies of the databases would
  be quite inefficient (it may be the case that the
  most of the records in both databases are the
  same).
• Databases are synchronized via comparison of
  sequence numbers.
• Interesting records - the sequence numbers are
  different or not present in database.
• The process is known as bringing up adjacencies.

61
Link State (LS) versus Distant Vector (DV)
Protocols

• In DV send everything you know to your neighbors.
• In LS send info about your neighbors to everyone.
• Msg size small with LS, potentially large with
  DV
• Msg exchange LS O(nE), DV only to neighbors

62
LS vs. DV

• Convergence speed
• LS fast
• DV slow, but faster with triggered updates.
• Space requirements
• LS maintains entire topology.
• DV maintains only neighbor state.
• Robustness
• LS can broadcast incorrect/corrupted LSP.
• localized problem
• DV can advertise incorrect paths to all
  destinations
• Incorrect calculation can spread to entire
  network.

63
LS vs. DV

• In LS nodes must compute consistent routes
  independently - must protect against LSDB
  corruption.
• In DV routes are computed relative to other
  nodes.
• Conclusion no clear winner, but the
  recomendation is to use DV only in small network
  and to use LS in all other cases.