IP Routing: GGP and RIP

1
IP Routing GGP and RIP

• Network Protocols and Standards
• Autumn 2004-2005

2
IP Routing Protocols

• Autonomous System
• Interior Gateway Protocols
• GGP
• RIP
• OSPF
• Exterior Gateway Protocols
• BGP
• EGP
• IP Multicast Routing
• MPLS

3
IP Routing Protocols

• Autonomous Systems

4
Routing in the Internet

• Routing Algorithms
• Bellman-Ford
• Dijkstra
• Routing Protocols
• Distance Vector
• Link State
• Routing Hierarchy
• Interior Gateway Protocols (RIP, OSPF, IGRP)
• Exterior Gateway Protocols (EGP, BGP, CIDR,
  Policy Routing)
• Multicasting (IGMP)

5
Internet from the start

• First, there was ARPANET
• Routers had complete information about all the
  possible destinations core routers
• GGP (gateway-to-gateway) protocol was used for
  routing a distance vector protocol

R
R
H
R
H
R
H
6
Internet from the start

• Then, LANs were connected to ARPANET

7
Internet from the start

• Problems with above configuration
• Routing overhead increased with the number of
  connected routers
• Number of routes increased with the number of
  connected segments
• Frequency of routing exchanges increased
• Higher likelihood that something went wrong
  somewhere requiring updates
• Number of different types of routers increased
• Slow deployment of new versions of routing
  algorithms

8
Internet from the start
Backbone Network
R1
Core Router
Local Network
R2
R3
R4
Local Network
Local Network
Local Network
9
Autonomous System
Backbone Network
R
R
Core Routers
R
AS
AS
AS
AS Autonomous System
10
Autonomous System

• What is an autonomous system?
• A set of routers and networks under the same
  administration. Examples
• A single router directly connecting one local
  network to the Internet
• A corporate network linking several local
  networks through a corporate backbone
• A set of client networks served by a single ISP
• NOTE From a routing point of view, all parts of
  an AS must remain connected

11
Autonomous System

• Internal connectivity within the AS means
• All routers must be connected
• Parts of network connected through core AS (yes,
  core is an AS!) cannot form an AS
• All routers must exchange routing information in
  order to maintain the connectivity (normally
  achieved by using a single routing protocol)
• Routers inside an AS are called interior
  gateway and the protocol they use is called
  Interior Gateway Protocol (IGP)

12
Autonomous System

• In 1982, the IGP of choice was GGP
• IGPs in use today are
• RIP
• OSPF
• IGRP
• Each AS is identified by a 16-bit number
• Number is assigned by the numbering authorities

13
Autonomous System Benefits

• Routing overhead is lower
• Network management becomes easy
• Easier computation of new routes
• Distribution of new software versions is easier
• Failing elements can be isolated easily
• AS use an Exterior Gateway Protocol to exchange
  information about reachability

14
IP Routing Protocols

• Gateway-to-Gateway Protocol
• GGP

15
GGP

• The old ARPANET routing protocol
• Defined in RFC 823
• A distance-vector routing protocol
• Only core routers participate in GGP
• GGP messages travel in IP datagrams with protocol
  type 3
• GGP measures distance in router hops. i.e., the
  number of hops along a path refers to the number
  of routers

16
GGP Message Types

• 4 types of GGP messages
• GGP Routing Update message (type 12)
• GGP Acknowledgment message (type 2/10)
• GGP Echo Request or Reply (type 0 or 8)

17
GGP Routing Update

• A router sends this message to advertise the
  destination networks it knows how to reach
• To keep the size of message small, networks are
  grouped by distance
• In the message Distance is followed by a list
  of Net addresses that are at this distance
• Contains a field that tells how many distance
  groups are being reported (3 in case below)
• D1 Net1, Net5, Net11
• D2 Net4, Net2, Net7, Net16
• D3 Net6, Net9

18
IP Routing Protocols

• Routing Information Protocol
• RIP

19
Routing Information Protocol

• A distance vector based IGP
• Similar to GGP
• Designed at UC Berkeley
• Based on Xerox XNS
• Distributed with 4BSD UNIX (routed)
• First RFC was 1058, current RFC is 2453
• Started off in small networks and then extended
  to larger networks
• See Huitema, Chapter 5

20
RIP Details

• Routers are active machines
• Advertise their routes (IP NET, distance) to
  others
• Hosts are passive machines
• They listen and update their routes but do not
  advertise
• RIP uses hop count metric
• RIP messages are transmitted using UDP at port 520

21
RIP Route Computation

• There is a cost associated with each link
• Typically cost 1 i.e., number of hops
• Each router receives route advertisements from
  its neighbors
• Advertisements show distances to all destinations
  in the network
• For each destination in the network
• The router takes each received advertisement and
  adds to it the cost to reach that neighbor who
  sent this advertisement this gives the distance
  to the destination
• The router selects lowest of these as path/cost
  to that destination

22
Algorithm Properties

• Convergence is guaranteed in a finite time given
  that topology remains static
• Starting value of distance estimates to each
  destination can be any non-negative number
• No assumption is made as to when the updates are
  sent or when the distances are computed
• Each router can work based on its own clock and
  send its updates asynchronously
• If the network changes, routes converge to a new
  equilibrium point

23
Example
Router
Advertisement Distance to A is 2 Distance to B
is 3 Distance to C is 5
Advertisement Distance to A is 1 Distance to B
is 4 Distance to C is 1
Cost 3
Cost 1
P1
P3
P2
Cost 2
Advertisement Distance to A is 2 Distance to B
is 1 Distance to C is 3
Distance to Through
Destination Port P1 Port P2 Port P3
A 3 4 4
B 4 3 7
C 6 5 4
24
Counting to Infinity
Routes to Target A route via B, distance 3 B
route via D, distance 2 C route via B, distance
3 D direct, distance 1
Assume that B to D link goes down, and B notices.
To reach target
x destination unreachable di directly
connected
What if the link from C to D also goes down?
Counting to Infinity!!!
25
Some Solutions

• Split Horizon
• If A reaches a destination through B, it makes no
  sense for B to reach the same destination through
  A
• Instead of broadcasting the same distance vector
  on all links, send different versions on each
  outgoing link by removing the entries for the
  destinations that are reachable through that link
• Split Horizon with Poisonous Reverse
• Include all the destinations in advertisements
  even those which were missing in split horizon,
  but
• Set those vector distances to infinity that were
  missing in the simple version of split horizon

26
Triggered Updates

• Split Horizon can work in loops with two
  gateways, but not with three or more
• See example in book by Huitema
• Another solution to deal with count to Infinity
  problem is triggered updates
• A gateway is required to send an immediate update
  when any route changes. This reduces the
  occurrence of loops
• Flood of triggered updates resolves loops faster
  when these happen

27
RIPv2 Message Format
16
8
24
31
AS NUMBER
VERSION (2)
COMMAND (1-5)
AUTHENTICATION TYPE
FFFF
AUTHENTICATION HEADER
MUST BE ZERO
FAMILY OF NET 1
ADDRESS OF NET 1
MASK
NEXT HOP
DISTANCE TO NET 1

28
Message Format
Command Meaning
1 Request for partial or full routing information
2 Response containing network-distance pairs from senders routing table
3 Turn on trace mode (obsolete)
4 Turn off trace mode (obsolete)
5 Reserved for Sun Microsystems Internal Use
29
RIPv2 Message Format

• Address format is not limited to TCP/IP
• RIP can be used with multiple network protocol
  suites
• Family of net i
• Identifies the protocol family under which the
  network address should be interpreted
• IP addresses are assigned value 2
• Next hop
• The sending router can specify another routers
  IP address as next hop for the network
• Set to 0.0.0.0 for sender itself
• Solves similar problem (extra hop) as ICMP
  redirect

30
RIP Metrics and Updates

• By default, RIP uses hop count as the distance
  metric
• Integers 1 through 15
• 16 denotes infinity
• Packets are normally sent every 30sec
• If a route is not refreshed within 180 seconds,
  distance is set to infinity and later entry is
  removed

31
Input Processing

• How to process incoming RIP packets?
• Examine entries one by one
• Validation check
• Address is valid class A, B, or C
• Network number is not 127
• Host port is not a broadcast address
• Metric is not larger than infinity (16)
• Incorrect entries are ignored
• And should be reported as errors

32
Input Processing

• Metric for entry is increased by link cost
• Routing table is searched for an entry
  corresponding to the destination
• If the entry is not present, it is added
• If the entry is present but with a larger metric
• Entry is updated and timer restarted
• Entry is present and next hop router is sender of
  response message
• Metric is updated and timer restarted
• For all other cases, entry is ignored

33
RIP Responses

• A separate response is prepared for all connected
  interfaces/ports
• Information sent on different ports may vary due
  to
• Split Horizon processing
• Subnet summarization
• For triggered updates may include only those
  entries that have been updated since last
  transmission
• Maximum message size 512 bytes (up to 25
  entries)
• Multiple messages have to be sent if more than
  512 bytes
• Source IP address is that of the interface on
  which the message is sent
• Destination IP address is the broadcast address