Ch06 Routing

McGraw-Hill The McGraw-Hill Companies, Inc., 2000

6-1

Chapter 6

Delivery and Routing of IP Packets


6-2

The network layer protocol first makes a connection between the source and destination hosts, then starts transmission of data. After the transmission of data is over, the network layer tears down the connection.

The network layer protocol treats each packet independently, with each packet having no relationship to any other packet.

Connection-Oriented Services

Connectionless Services


6-3

Direct delivery

Hosts are on the same (physical) network (no routing involved)


6-4

Indirect delivery

Hosts are not on the samenetwork (routing necessary)


6-5

Routing Methods:

! Next-Hop Routing! Network-Specific Routing! Host-Specific Routing! Default Routing


6-6

Next-hop routing

Only the next router is specified in the RT for each destination host


6-7

Routing is the relaying (forwarding) of packets generated by a source station, across the network in order to deliver them to their destination station. When a routing node gets a packet, it should know which neighboring node to for-ward the packet.

All routing strategies optimize some kind of performance criteria. This can be a minimal number of intermediate links along the path (minimum hop criteria), or a minimum cost criteria, which minimizes the total cost along the route (cost could be delay associated with each link, the inverse of the link's bit rate, etc.) Note that the minimal cost criteria is a generaliza- tion of the minimum hop, i.e. if we assign a cost = 1 to each link we get the minimum hop criteria.

The optimal routes from a set of source stations to corresponding destination stations are defined through routing tables (RT). Each node has a routing ta-ble which tells how a node can forward a package with a specified destinationtion.

(The following seven slides explain the essence of the next-hop routing)


6-8

Suppose that the costs associated with the internode links are as shown. (The minimal cost routes are shown on the next slide).

A

B C

D

E

F

G

1 2

3

5

46

3

2

15

7

1

4


6-9

1 2

3

5

46

1 2

3

5

46

Dest. Nextnode node

1 2 3 4 5 6

-26266

RT at node 1

Dest. Nextnode node

1 2 3 4 5 6

1-4441

RT at node 2

A

B C

D

E

F

G

1 2

3

5

46

3

2

15

7

1

4


6-10

A

B C

D

E

F

G

1 2

3

5

46

3

2

15

7

1

4

1 2

3

5

46

1 2

3

5

46

Dest. Nextnode node

1 2 3 4 5 6

11555-

RT at node 6

Dest. Nextnode node

1 2 3 4 5 6

223-55

RT at node 4


6-11

A

B C

D

E

F

G

1 2

3

5

46

3

2

15

7

1

4

1 2

3

5

46

1 2

3

5

46

Dest. Nextnode node

1 2 3 4 5 6

6434-6

RT at node 5

Dest. Nextnode node

1 2 3 4 5 6

54-455

RT at node 3


6-12

1 2

3

5

46

Dest. Nextnode node

1 2 3 4 5 6

-26266

RT at node 1

"An optimal policy has the property that whatever the initial state and the initial decision are, the remaining decisions must constitute an optimal policy with regard the state resulting from the first decision"

Is it suffitient to give in RT the address of the next node only instead of the entire route? For example, the first diagram contains five optimal routes emanating from the node "1": Source: 1, destination: 2, optimal route: 1,2 Source: 1, destination: 3, optimal route: 1,6,5,3 Source: 1, destination: 4, optimal route: 1,2,4 Source: 1, destination: 5, optimal route: 1,6,5 Source: 1, destination: 6, optimal route: 1,6

What justifies this?

This is possible because of the principle of optimality (stated by Richard Bellman in 1950s):


6-13

This can be applied to the optimal routes:

If the route a-x-b is optimal, then the remaining route x-b is optimal too, i.e. we would obtain exactly the same optimal route x-b if we started only from x. Therefore, the node a doesn't have to worry about the entire path, its is enough that it routes the packet to the optimal neighboring node x. This saves a lot of memory, because each node need to store only the optimal route to the next (neighboring) node.

a

b

First link ofthe optimalroute a-x-b

Remaining part of the optimal

route a-x-b

x


6-14

Network-specific routingOnly one entry exists in RT for all hosts connected to the same physical network

(the most common method)

Imagine 1000 hosts connected to the same network


6-15Host-specific routingLess efficiency, but the administrator has more control over routing.

Example: administrator wants to rout specifically the packets for B via routerR1 instead of R3, then the RT for A needs an entry: Host B " R1.This entry should precede the entry N3 " R3


6-16

Default routingAnything not

specified above


6-17

A static routing table contains information entered manually.

A dynamic routing table is updated periodically using one of the dynamicrouting protocols such as RIP, OSPF, or

BGP.

Static vs. Dynamic Routing

Used in small internetworks that do not change very often

Used in large internetworks that change very often. Automatic updates of RT makes the network manageable and scalable.


6-18

Routing module and routing table

Receiving the packet from the link layer and its basic processing (see chapter 8)

Packet has DA (besides SA), which is not necessarily equal to the next-hop address(see next slide)


6-19

HostA

RouterR

RouterS

HostB

SA=A, DA=B, payloadIP datagram:

Next hop: Router R


Next hop: Router S


Next hop: direct delivery

IP datagram stays basically the same all the way (exception: TTL)


6-20Routing Table

U The router is up and running. G The destination is in another network (Gateway).H Host-specific address. D Added by redirection (see chapter 9).M Modified by redirection (see chapter 9).

Number of users that connected to this destination via this router

Number of packets transmitted to this destination via this router

Interface name (like eth0, fddi0), identifies the port


6-21Routing Table (cont.)Routing table of a host can be displayed by commands:

route PRINT (Windows)netstat nr (UNIX)

Windows has a specific way of displaying the routing table: 1st line shows the default router, the 2nd line is loopback route, the 3rd line defines the range of addresses on the local network segment, the line with 224.0.0.0 is for multicast addresses, the line with 255.255.255.255 is for broadcast.


6-22

Get the packet;for (each entry in RT) {

Apply the mask to packet destination address;if (the result matches the value in destination field) {

if (G flag present)Use the next-hop address from the RT;(indirect delivery)

elseUse the destination address from the packet;(direct delivery)

Send the packet to the fragmentation module;return; // Routing done

}} // No match is foundSend an ICMP error message; return;

Routing Module


6-23

Configuration for routing example


6-24

Mask Destination Next Hop I/f

255.0.0.0 111.0.0.0 -- m0

255.255.255.224 193.14.5.160 - m2

255.255.255.224 193.14.5.192 - m1

--------------------------------------------------------------------------------255.255.255.255 194.17.21.16 111.20.18.14 m0

---------------------------------------------------------------------------------

255.255.255.0 192.16.7.0 111.15.17.32 m0

255.255.255.0 194.17.21.0 111.20.18.14 m0

---------------------------------------------------------------------------------

0.0.0.0 0.0.0.0 111.30.31.18 m0


6-25Example 1Router R1 receives 500 packets for destination 192.16.7.14; the algorithm applies the masks row by row to the destination address until a match (with the value in the second column) is found:

Direct delivery

192.16.7.14 & 255.0.0.0 # 192.0.0.0 no match

192.16.7.14 & 255.255.255.224 # 192.16.7.0 no match

192.16.7.14 & 255.255.255.224 # 192.16.7. no matchHost-specific

192.16.7.14 & 255.255.255.255 #192.16.7.14 no match

Network-specific

192.16.7.14 & 255.255.255.0 #192.16.7.0 match


6-26Example 2

Router R1 receives 100 packets for destination 193.14.5.176; thealgorithm applies the masks row by row to the destination address until a match is found:

Direct delivery

193.14.5.176 & 255.0.0.0 # 193.0.0.0 no match

193.14.5.176 & 255.255.255.224 #193.14.5.160 match


6-27Example 3

Router R1 receives 20 packets for destination 200.34.12.34; the algorithm applies the masks row by row to the destination address until a match is found:

Direct delivery

200.34.12.34 & 255.0.0.0 #200.0.0.0 no match

200.34.12.34 & 255.255.255.224 #200.34.12.32 no match

200.34.12.34 & 255.255.255.224 #200.34.12.32 no match

Host-specific

200.34.12.34 & 255.255.255.255 #200.34.12.34 no match


6-28Example 3 (cont.)

Router R1 receives 20 packets for destination 200.34.12.34; the algorithm applies the masks row by row to the destination address until a match is found:

Network-specific

200.34.12.34 & 255.255.255.0 # 200.34.12.0 no match

200.34.12.34 & 255.255.255.0 # 200.34.12.0 no match

Default

200.34.12.34 & 0.0.0.0 # 0.0.0.0. match


6-29

Example 4Make the routing table for router R1 in this figure:

Mask Destination Next Hop I.255.255.0.0 134.18.0.0 -- m0

255.255.0.0 129.8.0.0 222.13.16.40 m1

255.255.255.0 220.3.6.0 222.13.16.40 m1

0.0.0.0 0.0.0.0 134.18.5.2 m0


6-30

Example 5 Make the routing table for router R1 in this network:

Mask Destination Next Hop Interface255.255.255.0 200.8.4.0 ---- m2255.255.255.0 80.4.5.0 201.4.10.3 (or 200.8.4. 12) m1 (or m2)255.255.255.0 80.4.6.0 201.4.10.3 (or 200.4.8.12 ) m1 (or m2)0.0.0.0 0.0.0.0 ????????? m0


6-31

Example 6

Mask Destination Next Hop Interface255.255.0.0 110.70.0.0 - m0255.255.0.0 180.14.0.0 - m2255.255.0.0 190.17.0.0 - m1255.255.0.0 130.4.0.0 190.17.6.5 m1255.255.0.0 140.6.0.0 180.14.2.5 m20.0.0.0 0.0.0.0 110.70.4.6 m0

The routing table for router R1 is given below. Draw its topology


6-32

Routing Table Size

Hierarchical Routing

Geographical Routing

RT Search Algorithms

Classless Addressing and CIDR(CIDR = Classless Inter Domain Routing)

The size of RT is generally increasing(division of classes A and B into smaller blocks increase the number of RT entries.) Only the supernetting would decrease the number of entries (grouping of C blocks).

Way to decrease the size of RTs

Further decrease of RT size(large blocks: North America, Europe, Asia, Africa,)

RT is divided in buckets in order to localize the table search (in classful addressing the buckets are classes, in classless addressing the buckets are different prefix lengths. First is searched the bucket with the largest prefix = /32 (the host-specific routing), then the next prefix /31 is searched, etc.


6-33

Internet today

ISP = Internet Service ProviderNational ISPs (backbone networks):

SprintLink, PSINet, UUNet Technology, AGIS, Internet MCI

NAP = Network Access PointInternet interconnection points, typically work with ATM at speeds

DS-3 (T3) 44.736 MbpsOC-3 155.52 MbpsOC-9 466.56 MbpsOC-12 622.08 MbpsOC-48 2488.32 Mbps

Local ISPs


6-34Hierarchical Routing

Way to keep RTs smaller and more manageable

X Y Z

A AX1 Z2Y2X2 Y1 Z1X3

YXZ

X1 X2 X3 Y1 Y2 Z1 Z2

R2

R1

R3R4

Router R2 can see only networks X1, X2, X3


6-35

Hierarchical Routing (cont.)

X Y Z

A AX1 Z2Y2X2 Y1 Z1X3

YXZ

X1X2 X3 Y1 Y2 Z1 Z2

This part of the network doesnt have to be aware

of the division in the network X

Ch06 Routing

Documents

Transcript of Ch06 Routing