Evolution of Routing Techniques

A Presentation by
Tusharadri Sarkar
Evolution of Routing
Techniques
December 7, 2009Tusharadri Sarkar1

Delivery, Forwarding, Routing
IP and Mask
Routing Table
Unicast Routing Protocols
AGENDA

What is routing?
Why is routing required?
At which layer is routing done?
How does a router work?
Some Basic Questions

What is routing?
Routing is the process of selecting a path in
a network along which the packets shall be
sent to a destination
Routing consists of
A Router
A set of routing protocols
A routing information base (RIB)
One or more routing algorithms

Why is routing required?
 For practical limitation of physical
connections
 For efficiently managing the network traffic
 For efficient usage of network resources
 For catering to different types of services
 For congestion control

At which layer is routing done?
 Generally routing is done at network layer
 Multilayer layer routing and Cross layer routing is also
prevalent nowadays
 Firewalls are often integrated with routers

Router, Switch and Hub
The basic difference is varying intelligence

How does a router work?

Delivery
The network layer supervises the
handling of packets by the underlying
physical network
Every packet undergoes at least one
“Direct Delivery” and one or more
“Indirect Delivery”

Direct and Indirect Delivery
To rest of Network
Source SourceDestination
Destination
Direct
Indirect
Direct
Direct

Logical addressing: IP and MASK
Internet Protocol address is a logical and global
addressing scheme
It uniquely defines the connection of a
device/network to the Internet
IPv4: 32 bit addressing scheme
Address space: 232
= 4294967296
Notations: Dotted Decimal:
117.149.29.2
Notations: Binary:
01110101 1001010 00011101 00000010

 Classful addressing
1st
2nd
3rd
4th
0
10
110
1110
1111
1st
2nd
3rd
4th
0-127
128-191
192-223
224-239
240-255
Classes
Class A
Class B
Class C
Class D
Class E
Class No. of Blocks Block Size Application
A 128 16,777,216 Unicast
B 16,384 65,536 Unicast
C 2,097,152 256 Unicast
D 1 268,435,456 Multicast
E 1 268,435,456 Reserved

Mask: A 32 bit number made of n contiguous 1s
followed by (32-n) contiguous 0s (n<32)
Default masks for Classful addressing:
Given an IP and its mask, one can calculate:
First Address
Last Address
Range of Addresses
Class Binary Dotted Decimal CIDR
A 11111111 00000000 00000000 00000000 255.0.0.0 /8
B 11111111 11111111 00000000 00000000 255.255.0.0 /16
C 11111111 11111111 11111111 00000000 255.255.255.0 /24

Classless addressing: No more classes but a
block of addresses are assigned, provided the
following restrictions are strictly followed
The addresses in the block must be contiguous
The number of addresses must be a power of 2
The first address must be evenly divisible by the total
number of addresses allocated
Mask is a better way to define a block
An example: Given an IP address
205.16.37.39/28
What are the first, last and the total number of
addresses assigned?

Binary equivalent of mask /28:
 11111111 11111111 11111111 11110000 (255.255.255.240)
Binary equivalent of the address:
 11001101 00010000 00100101 00100111 (205.16.37.39)
First address: Set the right most 4 bits to 0:
 11001101 00010000 00100101 0010000 (205.16.37.32)
Last address: Set the right most 4 bits to 1:
 11001101 00010000 00100101 00101111 (205.16.37.47)
Number of addresses: 232-n
= 24
=16
So, in general a address in classless addressing is
mentioned as: x.y.z.t/n

Network Address
 When a organization is allocated a block of addresses,
normally (not always) the first address is treated as the
network address
 It is not assigned to any device, it defines the organization
itself to the rest of the world
REST of the WORLD
Network Address:
205.16.37.32
All packets with receiver
address 205.16.37.32 to
205.16.37.47 are routed to
x.y.z.t/nx.y.x.t/n 205.16.37.32/28
205.16.39.33/28 205.16.39.47/28
… …

Routing Table
A host or a router maintains a ‘routing table’ with
an entry for each specific destination
The table can be STATIC or DYNAMIC
Static Routing Table:
Contains information entered manually by the
administrator at the time of creation
Cannot be modified automatically when there is any
change in the Internet
Dynamic Routing Table:
Capable of updating the table with the help of routing
protocols and algorithms automatically
Only option for managing any large network of today

Routing Table
Mask: Defines the mask applied to that entry
Network Address: Defines the network address to
which the packet is finally delivered. In host specific
routing, this is the destination host address
Next Hop Address: Defines the address of the hop
for the packet
Interface: Shows the name of the interfaces
Mask
Network
Interface
Next-hop
address
Interface Flags
Reference
Count
Use
… … … … … … …

Routing Table
FLAGS: Defines up to five flags
 U (Up): Router is up and running
 G (Gateway): Destination is in another network
 H (Host-specific): Network address is host-specific address.
Otherwise the network address is the destination address
 D (Added by redirection): Routing info is added to host routing table
by redirection message from ICMP
 M (Modified by redirection): Routing info for destination is modified
to host routing table by redirection message from ICMP
Reference Count: Defines number of users at this
route at the moment
Use: Defines number of packets transmitted through
the router for a destination

A quick look at a system routing table

Network Configuration of a System
form the Routing Table
A UNIX server gives the following result with netstat
and ifconfig command
 $ netstat –nr
 Kernel IP routing table
 $ ifconfig eth0
 Eth0 Link encap:Ethernet Hwaddr 00:B0:D0:DF:09:5D
 Inet addr: 153.18.17.11 Bcast: 153.18.31.255 Mask:255.255.240.0
What is the network configuration of the server?
Destination Gateway Mask Flags Iface
153.18.16.0 0.0.0.0 255.255.240.0 U eth0
127.0.0.0 0.0.0.0 255.0.0.0 U lo
0.0.0.0 153.18.31.254 0.0.0.0 UG eth0

Network Configuration from the
Routing Table
Rest of the Internet
153.18.16.0/20
153.18.31.254/20
153.18.17.11/20
eth0
00:B0:D0:DF:09:5D
Default
Router

Forwarding
It means placing the packet in its route to its
destination
Requires a host or a router to have a routing table
When the host has a packet to send or the router
has received a packet, it looks up this routing
table to determine route to the final destination
Routing techniques caters to optimizing this table
as maintain a full-fledged look-up table is
impossible to maintain

Forwarding Techniques
Next-hop method Vs Route Method
N1
N2 N3
R1 R2
Host A
Host B
Routing tables based on routing
Destination Route
Host B R1, R2, host B
Destination Route
Host B R2, host B
Destination Route
Host B Host B
Routing tables based on Next-hop
Destination Route
Host B R1
Destination Route
Host B R2
Destination Route
Host B Host B
For
A
For
R1
For
R2

December 7, 2009
Network-Specific method Vs Host Specific method
System
DCBA
N2N1
Routing table for host S based on
host-specific method
Destination Next Hop
A R1
B R1
C R1
D R1
Routing table for host S based on
network-specific method
N2 R1
R1
Tusharadri Sarkar25

Default Method: Using a default router
N1 N2
R1
R2
Host A
Rest of the Internet
Default
Router
N2 R2
Any other R1
Routing
table for
host A

Forwarding Process
 In classless addressing, at least 4 columns are required
 The routing table is searched based on the network address
and mask
Mask Network
Address
Next-hop
Address
Interface
… … … …
… … … …
… … … …
Extract
Destination
Address
Search
Table
Forwarding Module
To ARP
Next –hop address
and interface no.

Managing Routing Table in Classless
Addressing
 Address aggregation: Blocks of addresses of different
interface and mask are aggregated into one single block in
routing table
 Several levels of aggregation are possible
140.24.7.0/26
140.24.7.64/26
140.24.7.128/26
140.24.7.192/26
Org 1
Org 2
Org 3
Org 4
m0
m1
m2
m3
m4 m0 m1
R1 R2

Managing Routing Table in Classless
Addressing
Address aggregation: Routing tables for router R1
and router R2
For R2, any packet with destination addresses 140.24.7.0 to
140.24.7.255 are sent to interface m0 regardless of any of
the organizations
Mask NA NHA Iface
/26 140.24.7.0 … m0
/26 140.24.7.64 … m1
/26 140.24.7.128 … m2
/26 140.24.7.192 … m3
/0 0.0.0.0 Default m4
Mask NA NHA Iface
/24 140.24.7.0 … m0
/0 0.0.0.0 Default m1
Routing table for R1 Routing table for R2

Longest Mask Matching
 What happens if Org. 4 is not geographically close to the other
3 Orgs?
 Can we still use Address Aggregation and assign the block
140.24.7.192/26 to Org. 4?
R2
R1
R3
140.24.7.0/26
140.24.7.64/26
140.24.7.128/26
140.24.7.192/26
Org 1
Org 2
Org 3
Org 4
m0
m1
m2
m3
m0 m2
m1
m0m1
m2

Longest Mask Matching
 Answer: YES
 Reason: LONGEST MASK MATCHING
 The “Routing Table” is sorted from the longest mask to the
shortest mask
Mask NA NHA Iface
/26 140.24.7.0 … m0
/26 140.24.7.64 … m1
/26 140.24.7.128 … m2
/0 0.0.0.0 Default m3
Mask NA NHA Iface
/26 140.24.7.192 … m0
/0 0.0.0.0 Default m2
Mask NA NHA Iface
/26 140.24.7.192 … m1
/24 140.24.7.0 … m0
/0 0.0.0.0 Default m2
Routing table for R1 Routing table for R2
Routing table for R3 December 7, 2009Tusharadri Sarkar31

Hierarchical Routing
Hierarchical routing can greatly minimize the size of
the routing tables
For example, a regional ISP is granted a 16,384 (214
)
addresses starting from 120.14.64.0/18
It is divided in to 4 sub-blocks each of size 4096 for
3 local ISPs. For them the mask is /20
1st
local ISP divides its assigns sub-blocks into 8
smaller blocks for small ISPs. For them the mask
becomes /23
Each small ISPs divides them into 128 sub-blocks
for households. For them the mask becomes /30,
and so on…

Hierarchical Routing
 The logical representation is displayed here
120.14.64.0/18
Total 16,384
120.14.64.0/20
120.14.64.0/23120.14.64.0/30
120.14.78.0/30
120.14.78.0/23
120.14.80.0/20
120.14.96.0/22
120.14.112.0/24
120.14.96.0/20
120.14.112.0/20
Total
4096
Total
4096
Total
4096
Total
4096
512
512
ISP 1
ISP 2.1
ISP 3.1
ISP 3.8
ISP 2.2
ISP 2.3
Total 4 Large Orgs.
Total 16 Small Orgs.
128 Each
128 Each

Geographical Routing
The same concept of hierarchical routing can be
extended in geographical routing
To decrease the size of the routing tables further,
segregation is done in geographical level as well
For example, the entire address space is divided
into few large blocks
One block is assigned to North America, one to
Asia, one to Africa, one to Europe and so on…
So, for all the routers of the ISPs outside Europe,
every router will have one and only entry for all the
addresses assigned to Europe

Unicast Routing Protocols
Routing protocols are needed to maintain and
update dynamic routing tables
A routing protocols is a combination of set of rules
(algorithms) and procedures
Unicast routing protocols applies where each
incoming packet has to be delivered to one and
only one destination
Router decides the next hope of a packet in a
‘Autonomous System’ based on ‘Optimization’
3 most popular and basic Unicast Routing
Protocols are: RIP (Distance Vector routing), OSPF
(Path Vector routing) and BGP (Link State routing)

Autonomous Systems
An ‘Autonomous System’ or ‘AS’ is group of
networks and routers under the authority of a single
administration
Routing inside AS : Intra-domain routing
Routing between AS : Inter-domain routing
AS1
AS4AS3
AS2

Routing Protocols Scope

Optimization
The router must always choose the optimum path
between two networks for the packets
There is a cost associated with each packet for
passing through a network, called ‘Metric’
The metric is different depending on the routing
protocols. For example:
 In RIP, the hop count is used as the metric
 In OSPF, the administrator can assign a cost for
network based on the type of service required
 In BGP, the administrator can set the cost based on
the policy of the network

Distance Vector Routing
In DVR, the least cost route between any two nodes
is the route with minimum distance
Each node maintains a vector (table) of minimum
distance to every node known
There are three steps involved:
 Initialization: At the beginning, each node knows the
distance to its immediate neighbors
 Sharing: Periodically or in triggered time, the nodes
share their vectors with other nodes
 Updating: Based on the shared info, nodes updates
their vectors about path to indirectly connected
nodes

DVR: Initialization
To Cost Next
A 0 _
B 5 _
C 2 _
D 3 _
E ∞
A
C
B
ED
5
3
2 4
34
To Cost Next
A 3 _
B ∞
C ∞
D 0 _
E ∞
To Cost Next
A 5 _
B 0 _
C 4 _
D ∞
E 3 _
To Cost Next
A ∞
B 3 _
C 4 _
D ∞
E 0 _
To Cost Next
A 2 _
B 4 _
C 0 _
D ∞
E 4 _
Table of “A” Table of “B”
Table of “C”
Table of “D” Table of “E”

DVR: Sharing and Updating
Each node will share its routing table on periodic
basis or triggered condition
Full routing table needed not be shared. In our
scenario, only column 1 and column 2 will be shared.
Next Hop Address (column 3) will be calculated
based on that
Receiving a partial table from its neighbor, a node
calculates a temporary updated table
Then each row of the old and new table are
compared based on the next node entry (col. 3)
If next node entry is different, the row with smaller cost is
chosen. If there is a tie, old entry is kept
If next node entry is same, the new entry is chosen

DVR: Updating Table for “A”
To Cost Next
A 0 _
B 5 _
C 2 _
D 3 _
E ∞
To Cost
A 2
B 4
C 0
D ∞
E 4
To Cost Next
A 4 C
B 6 C
C 2 C
D ∞ C
E 6 C
To Cost Next
A 0 _
B 5 _
C 2 _
D 3 _
E 6 C
Old Table of “A”Table
Received
from “C”
Modified
Table of “A”
New Table of “A”
Compare

DVR: The Finalized Tables
To Cost Next
A 0 _
B 5 _
C 2 _
D 3 _
E 6 C
A
C
B
ED
5
3
2 4
3
4
To Cost Next
A 3 _
B 8 A
C 5 A
D 0 _
E 9 A
To Cost Next
A 5 _
B 0 _
C 4 _
D 8 A
E 3 _
To Cost Next
A 6 C
B 3 _
C 4 _
D 9 C
E 0 _
To Cost Next
A 2 _
B 4 _
C 0 _
D 5 A
E 4 _
Table of “A” Table of “B”
Table of “C”
Table of “D” Table of “E”

DVR: Two Node Loop Instability
X A B
XX
X
X
A
AA
A
B
B
B
B
2 4
4
4
4
4
.
.
.
X 2 _
X ∞ _
X 10 B
X 10 B
X ∞ _
X 6 A X 14 A
X 6 A
X 6 A
X ∞ _
Before
Failure
After
Failure
After A
receives
update
from B
After B
receives
update
from A
Finally

DVR: Preventing Instability
Defining ‘INFINITY’: Infinity should be defined as a
smaller number say, 100. In RIP ‘Infinity’ is often
defined as 16. So, the network can’t have more than
15 hops anywhere.
Split Horizon: Each node sends only part of its table
through each interface. In our case, B would not
advertize its part of the table to A which contains
information about X (i.e. the route of X is through A,
so A already knows).
Split Horizon & Poison Reverse: While sharing its
table with A, B will add a tag to the route information
of X (i.e. “I know this route comes from you. Please
do not use this value”).

Routing Information Protocol
Routing Information Protocol (RIP) is an
implementation of ‘Distance Vector Algorithm’ with
the following considerations:
1. In an autonomous system, we are dealing with
routers and networks (links). Only routers have
routing tables, networks not
2. The destination in a routing table is a network
always
3. The metric used by RIP is the no of hops
needed to reach the destination
4. Infinity is defined as 16
5. The next-node column defines the address of
the router to which packet is to be sent

Link State Routing
Domain Topology: Here, each node in the domain has
an entire topology of the domain
Link State: For each node, the number of other links
and nodes, their connectivity type, cost (metric) and
the condition of the links (Up or Down) constitutes link
state
Shortest Path Tree: Based on the link states, a node
can use Dijkstra’s Algorithm to create a ‘Shortest Path
Tree’ which can used as the routing table
There are four sets of operations required
Creation of Link State Packets (LSPs)
Flooding of LSPs
Formation of shortest path tree
Calculation of routing based on the tree

Link State Routing:
A
C
B
ED
5
3
2 4
State of
Links
for “A”
 Initial Condition:
D
A B
E
5
2
3
3
2 4
4
5
4
3
34
3
State of
Links
for “D”
State of
Links
for “B”
State of
Links
for “E”
State of
Links
for “C”

Link State Routing:
 Dijkstra’s Algorithm: Formation of Shortest Path Tree
START
STOP
Tentative list
is empty?
Set root to local node and
move it to tentative list
Among nodes in tentative list, move the
ones with shortest path to permanent list
Add each unprocessed neighbor of last
moved node to tentative list if it is not there
already. If neighbor is in tentative list with
larger cumulative cost, replace with new one
YES
NO

Link State Routing:
A B
0
1. Set root to A and move A to tentative list
 Creation of Shortest Path Tree for node A:
A
Permanent List: Empty Tentative List: A(0)
Root

Link State Routing:
A B
0
2. Move A to permanent List. Add B, C, D to tentative list
A
Permanent List: A(0) Tentative List: B(5), C(2), D(3)
Root
5B
2 C
3 D

Link State Routing:
A B
0
3. Move C to permanent List. Add E tentative list
A
Permanent List: A(0), C(2) Tentative List: B(5), D(3), E(6)
Root
5B
2 C
3 D 6E

Link State Routing:
A B
0
4. Move D to permanent List.
A
Permanent List: A(0), C(2), D(3) Tentative List: B(5), E(6)
Root
5B
2 C
3 D 6E

Link State Routing:
A B
0
5. Move B to permanent List.
A
Permanent List: A(0), B(5), C(2), D(3) Tentative List: E(6)
Root
5B
2 C
3 D 6E

Link State Routing:
A B
0
6. Move E to permanent List.
A
Permanent List: A(0), B(5), C(2), D(3), E(6) Tentative List:
Empty
Root
5B
2 C
3 D 6E

Link State Routing:
 Calculation of Routing Table from Shortest Path Tree
Node Cost Next
A 0 _
B 5 _
C 2 _
D 3 _
E 6 C
Routing table for node A
 We can see that the
routing table of A as
deduced by Link State
Routing is the same as
Distance Vector Routing
 In real scenario, the
routing table is
determined by the cost
assigned to each node by
the administrator

Open Shortest Path First (OSPF)
OSPF is based on Link State Routing Protocol
Area: A collection of networks, hosts and routers all
contained within an autonomous system
Area Border Routers: Summarizes all the information
about an area and shares it across
Backbone: A special area among all areas in an AS
which all other areas must be connected to. The
backbone always has area code ‘0’
Backbone Routers: Routers in a backbone. A
backbone router can also be area border router
Virtual Link: If the connection between an area and
backbone is broken the administrator can create an
alternate connection between routers

OSPF: Implementation
net net net
net
net
net
net
net net
net
Area 1
Area 2
Area 0 (Backbone)
ABRABR
BR
BR
AS BR
Autonomous System (AS)

Path Vector Routing
Why DVR and LSR are not suitable for inter-domain
routing?
 Reason: Scalability
DVR becomes instable and intractable for a large
number of hops (even more than 16)
LSR needs a huge amount of resource to calculate
its shortest paths. It also causes heavy traffic in the
network because of flooding of LSP
How path vector routing eliminates them?
Well, it is simply derived from DVR, but does not
assign hop count as the metric/cost...

Path Vector Routing
Speaker node: In path vector routing, a special
node acts on behalf on the entire AS. It summarizes
all the information of that AS, creates a routing
table and advertizes it to other ASs
What is advertized?
Not the metrics but the paths in an AS
Policy: Every AS will have a well defined policy
Paths are decided upon by the speaker nodes by
consulting the policies in neighboring ASs
Reason: Different ASs will have different policies &
priorities associated with them

Path Vector Routing
Initialization: At the beginning each SN knows only
about all other nodes inside its AS
Sharing: Just as in DVR, the speaker nodes will
then share their tables with immediate neighbors
periodically or on trigger
Updating: On receiving a two column table from
neighbor, a speaker node will update its table by
adding the nodes not present in its routing table as
well as adding its own AS and other ASs that sent
the table
Loop Prevention
Policy Routing
Optimum Path

Path Vector Routing
A1
D1
C1
B1
A2 A3
A4 A5
C2 C3
B2
B3
B4
D2 D3
D4
Dest. Path
A1 AS1
A2 AS1
A3 AS1
A4 AS1
A5 AS1
Dest. Path
C1 AS3
C2 AS3
C3 AS3
Dest. Path
B1 AS2
B2 AS2
B3 AS2
B4 AS2
Dest. Path
D1 AS4
D2 AS4
D3 AS4
D4 AS4
AS1
AS2
AS3
AS4
TableofA1
TableofD1
TableofC1
TableofB1

Path Vector Routing
 Stabilized tables of all AS after update:
Dest Path
A1
…
A5
AS1
…
AS1
B1
…
B4
AS1-AS2
…
AS1-AS2
C1
…
C3
AS1-AS3
…
AS1-AS3
D1
…
D4
AS1-AS2-AS4
…
AS1-AS2-AS4
Dest Path
A1
…
A5
AS2-AS1
…
AS2-AS1
B1
…
B4
AS2
…
AS2
C1
…
C3
AS2-AS3
…
AS2-AS3
D1
…
D4
AS2-AS3-AS4
…
AS2-AS3-AS4
Dest Path
A1
…
A5
AS3-AS1
…
AS3-AS1
B1
…
B4
AS3-AS2
…
AS3-AS2
C1
…
C3
AS3
…
AS3
D1
…
D4
AS3-AS4
…
AS3-AS4
Dest Path
A1
…
A5
AS4-AS3-AS1
…
AS4-AS3-AS1
B1
…
B4
AS4-AS3-AS2
…
AS4-AS3-AS2
C1
…
C3
AS4-AS3
…
AS4-AS3
D1
…
D4
AS4
…
AS4
Table of A1 Table of B1 Table of C1 Table of D1

Path Vector Routing
Some important features of updating:
Loop Prevention: The instability of DVR is avoided in
PVR; upon receiving a message the router checks to
see if its AS is in the path
Policy Routing: Upon receiving a message a router
checks the path with policy. If an AS in the path is
against policy it can ignore that
Optimum path: Router find the path that fits the
organization best. A path from AS4 to AS1 can either
be AS4->AS3->AS2->AS1 or AS4->AS3->AS1. Here
we will choose the path with less number of ASs
involved
This is not a general rule. There are complex criteria which
are always involved in real scenario

Border Gateway Protocol (BGP)
BGP was introduced in 1989
Some features of BGP:
Types of AS
 Stub AS: An AS which is connected to another AS. A
stub is either a source or a sink
 Multihomed AS: An AS which is connected to more
than one AS, but it is only a sink or source. Example:
A large corporation which is connected to more than
one regional or national ASs
 Transit AS: A multihomed AS that allows flow of data
traffic through it. Example: All national and
international ISPs

Path Attributes
 Well known attribute: Every BGP router must recognize
 Well known mandatory attribute: It must appear in the
description of a route; e.g. origin, next-hop
 Well known discretionary attribute: It must be recognized but
need not be included always in update
 Optional Attribute: Need not be recognized by all BGP
routers
 Optional transitive attribute: It must be passed to the next router
by the router that has not implemented it
 Optional non-transitive attribute: It must be discarded if the
receiving router has not implemented it

BGP Sessions: A BGP session is a connection setup
between two BGP routers for the sake of exchanging
router information
A session in BGP is a connection at the TCP level.
 External BGP Session (E-BGP): Takes place when two speaker
nodes exchange routing information
 Internal BGP Session (I-BGP): Takes place when a speaker
node collects information from other nodes in the its own As
A1
A2 A3
A4 A5
AS1
C1
C2 C3
AS3

Reference
December 7, 2009Tusharadri Sarkar
Data Communications and Networking
–Behrouz A. Forouzen
68

Thank You

Evolution of Routing Techniques

Recomendados

Recomendados

Más contenido relacionado

La actualidad más candente

La actualidad más candente (20)

Similar a Evolution of Routing Techniques

Similar a Evolution of Routing Techniques (20)

Último

Último (20)

Evolution of Routing Techniques