Subnet & Supernet

Chong-kwon Kim

SNU SCONE lab.

Problems of Addressing

IP addressing scheme is too rigid

– One network ID for each organization

– Only three classes

Problems

1. Large physical network (Large extended LAN)

2. Inefficient use of addresses

• Need to allocate Class B address to a network with 255 hosts

– 255/65535 = 0.39% efficient

SNU SCONE lab.

Subnetting-1

Solution for large organizations

A class A (or B) network may have tens of thousand of hostsProblems?

One solution is to assign many class C addresses Routing complexity increases

One Entry in

Forwarding Table

One

organization

One physical

network

One

organization

Many physical

networks

Many Entries in

Forwarding Table

Subnetting

Subnetting-2

Partition a large network into multiple small physical networks called subnet

Use a part of the host ID space for subnet identification– How do you know What part of

host ID is used for Subnet?

Subnet mask

Routing– Outside, route based on network ID (prefix) only

– Inside, route based on (network+subnet ID)

SNU SCONE lab. 4

SNU SCONE lab.

Subnet Example & Forwarding

ProcedureSubnet Mask: 255.255.255.128

Subnet Number: 128.96.34.0 (00100010 00000000)

Subnet Mask:

255.255.255.128

Subnet number:

128.96.34.128 (00100010 10000000)

Subnet Mask:

255.255.255.0

Subnet number:

128.96.33.0(00100001 00000000)

128.96.34.15 128.96.34.1

128.96.34.130

128.96.34.129 128.96.34.139

128.96.33.14 128.96.33.1

Router R0 Forwarding Table

SubnetNo SubnetMask NextHop

128.96.34.0 255.255.255.128 Interface 0

128.96.34.128 255.255.255.128 Interface 1

128.96.33.0 255.255.255.0 R1

R0

R1

IP Lookup procedure

Let D = Destination IP address

For each forwarding entry

D1 = SubnetMask & D

if D1 = SubnetNumber

Deliver to the NextHop

break

6

CIDR (Supernetting)

Solution to efficient use of address

Allocate multiple (small) network IDs to an organization such that they can be aggregated into one prefix

CIDR(Classless Inter-Domain Routing), Supernetting– Ignore IP address class

– Variable network ID length

– Prefix: Network ID part of IP addresses

1010…..11 00

1010…..11 01

1010…..11 10

1010…..11 11

1010…..11

Prefix (22 bits)

Scenario:

A company with 900 hosts

10 bit host ID

24 bit

CIDR - Example

Network ID allocation & Aggregation– For a network with N hosts, host ID length should be at least n

where 2^n > N

• Use 32 -n bits for network ID

– Example:

• For an organization with 4,000 hosts, host ID part should be 12 bit long

– Network ID part is 20 bit long

– Share a common prefix (network ID part) of desired length

– Example

• 192.4.16.0-192.4.31.0 (11000000 00000100 0001xxxx xxxxxxxx)

SNU SCONE lab. 7

SubnetNumber SubnetMask NextHop

128.96.34.0 255.255.255.128 Interface 0

128.96.34.128 255.255.255.128 Interface 1

128.96.33.0 255.255.255.0 R2

SNU SCONE lab. 8

CIDR Notation

Notation

– IPAddress / length

• Length specifies the prefix used for network ID

• Similar to subnet mask

• 185.21.16.0/20 = 255.255.240.0

Forwarding table entry

– Use prefix length instead of subnet mask

SubnetNumber NextHop

128.96.34.0/25 Interface 0

128.96.34.128/25 Interface 1

128.96.33.0/24 R2

9

CIDR & Routing Route aggregation

Longest matching prefix

Destination NextHop

128.112.128.0/24 Int 0

128.112.128.0/21 Int 1

Forward a packet to 128.112.128.*?

Forward a packet to 128.112.129.*?

ISP2

Router XAdvertise

128.112.128/24

Routing

Chong-kwon Kim

11

Routing & Forwarding

Routing– Collect network information and determine shortest paths

– Path selection criteria

• Hop count, distance, reliability, QoS, …

– As a result, generate forwarding tables

Forwarding– Move packets according to forwarding table

SNU SCONE lab.

SNU SCONE lab. 12

Routing Scalability

Millions of networks in the Internet

Scalability problem

– Control packet overhead

– Processing overhead

Divide and conquer, Abstraction

Partition the Internet into pieces called AS

(Autonomous system) or RD(Routing Domain)

– Single authority unit over

• Address management & Routing inside the domain

– Examples

• ISP, Large University/company, …

AS ID (16 bit 32 bit)

– Each AS has a unique ID

SNU SCONE lab. 13

Intra-/Inter-domain Routing Intradomain routing

– Routing within an AS where the owner has a complete control over the network operation

– Optimality > Reachability

– Collect all information & find shortest paths

– IGP (Interior Gateway Protocol)

– RIP, OSPF

Interdomain routing– Routing across AS boundaries

– AS would not disclose inside information

– Reachability > Optimality

– Exchange reachability info. between ASs

– EGP (Exterior Gateway Protocol)

– EGP, BGP-4

14

Graph Model Represent a network as a graph

– Node: network or router

– Link: network link

• Link cost

Find the shortest paths on the graph– Network conditions change dynamically

– Shortest path algorithms

Point-to-

point

Ethernet

FDDI

A

XY

Z B

x y z2 1 13

C=2

C=1

C=3

C=1

11

Ethernet

FDDI

P2P

Intradomain Routing Algorithm &

Protocol Routing algorithm = Shortest path algorithms

– Bellman-Ford algorithm

– Dijkstra algorithm

Routing protocol– Distributed realization of shortest path algorithms

• What information should be exchanged for distributeimplementation of shortest path algorithms?

– Application layer protocol that exchange

• Routing info.

• Network topology

• Network operating conditions

– Faults, congestion, estimated delay...

– RIP(Routing Information Protocol)

– OSPF(Open Shortest Path First)

SNU SCONE lab.

Shortest Path Algorithm

Graph G = (N,E)– Link (i, j) is incident on node i and j

• Associated cost,

– Path (i, j, k, l, ,m) is a series of links connecting two end nodes i and m

• Cost =

Shortest path algorithm– Find a path between two nodes with minimum cost

c ij

lmkljkij cccc

SNU SCONE lab.

Bellman-Ford Algorithm

Let D(v) be the cost of current shortest path from node v to s

Algorithm

Step 1: D(s) = 0

D(v) = for all v

Step 2: D’(v) = min [D(u) +

u ∈ N(v)

If D’(v) = D(v) for all v , Stop

O.W. D(v) = D’(v) for all v

Repeat

cuv ]

sv

l

m

n

Distributed Implementation?

x

y

SNU SCONE lab.

DV (Distance-Vector) Algorithm

Based on distributed BF Algorithm

Each node sends to neighbor nodes its own optimal path costs as

– Distance vector

• Shortest path cost to each destination

Each node receives distance vector from all of its neighbor nodes and compute best routes

sv

l

m

n 1

5

2

x

y

19

BF - ExampleDistance to network

Router

4 8 12 16

A 0 1 1

B 1 0 4

C 1 4 0 3

D 3 0

Router

4 8 12 16

A ? ? ? ?

B ? ? ? ?

C ? ? ? ?

D ? ? ? ?

192.168.

0.48 12 16

A 0

B 0

C 0

D 0

SNU SCONE lab. 20

RIP

Neighbor routers exchange RIP request/response

messages that contain distance vector

When to send RIP messages?

– Periodic

– Triggered

Command Version 0

Family of Net1 0

Distance to Net1

Subnet Mask

Next hop

IP Address of Net1

Net2

Net3

...

Use UDP Port 520

How to limit the delivery

range to directly connected

routers?

Route Adaptation

(F)

(A)

(D)

(G)

(A)

(G)

Next Hop

Suppose F notices that link (F, G) is broken① F advertises to A that its cost to G is

② A receives from B, C and E with cost = 3, 2, 3,

respectively

③ A updates its route to G via C with cost 3

④ F receives advertisement from A and updates

the route via A with cost 4

RIP Problem – Count-to-Infinity

SNU INC lab. 22

Suppose A notices that link (A, E) is broken① A advertises to B its cost to E is

② C advertises to B that its cost to E is 2

B updates its route to E via C with cost 3

③ A’s link failure message arrives to C

C think E is unreachable

④ B advertises to A that its cost to E is 3

A updates its route to E via B with cost 4

⑤ A advertises to C that its cost to E is 4

SNU SCONE lab. 23

RIP Problem

Slow convergence– Count to infinity

A

D

C

B

10

1

11

1

Routing Table Updates

A B C D

D NH D NH D NH D NH

2 B 1 D 2 B 0 dd

2 B ∞ ur 2 B 0 dd

3 C 3 C 3 A 0 dd

4 C 4 C 4 A 0 dd

Before Break

24

Split Horizon & Poison Reverse Horizon

– Directions where to advertise distance-vector

Split horizon

– Do not advertise a route to an interface from where the best

trigger (next hop) arrives

Split horizon with Poison reverse

– Advertise a route with ∞ to an interface from where the best

trigger arrives

A

D

C

B

101

11

1

Routing Table Updates

A B C D

D NH D NH D NH D NH

2 B 1 D 2 B 0 dd

2 B ∞ ur 2 B 0 dd

Read: http://technet.microsoft.com/library/Cc940478

SNU SCONE lab.

Dijkstra Algorithm

Find shortest paths from node s

d

3

s

a b

c

e

1

2

22

1

13

55

Algorithm

Step 1: F = {s}

D(s) = 0

D(v) = 𝐶𝑠𝑣

Step 2: If F = N, Stop

O.W. Find u s.t. D(u) = min { D(x) }

x ∈ N-F

F = F + {u}

D(v) = min [D(v), D(u) + 𝐶𝑢𝑣], ∀ v ∈ N(u)

Repeat

SNU SCONE lab. 26

Link State Routing Protocol - 1 Problem of distance-vector routing protocol

– Large overhead

– Slow convergence

– Not scalable

• No hierarchy

Search for a new intra-domain routing protocol

– Starts in 1987

– Multiple paths between a source-destination pair

– Descriptive metric

– Hierarchy

Distributed database model

– Each router maintains complete network information

SNU SCONE lab. 27

Link State Routing Protocol - 2 Dijkstra (or any SP algorithms)

Procedure– Each router monitors the status of directly connected links

– Announce the link state (& network) information to all routers

• LSA (Link State Advertisement)

• Use flooding

– Collect LSAs into the local link state database and compute the shortest path tree rooted at the router

LSA contains– ID of the node that creates the LSA

– Directly connected neighbor (router or network) and its cost(state)

SNU SCONE lab. 28

OSPF - Area

OSPF is very complex protocol

– Hierarchy

• A large AS/RD is partitioned

into several areas

– Load balancing

A large AS has thousands of routers

- Hierarchical structure

Area: a set of routers that exchange LSA

Area 0 : Backbone area

ABR(Area Border Router)

- Router that is both the member of

backbone area and non-backbone area

Use R4-R5 link?

Optimality vs. Scalability

SNU SCONE lab. 29

OSPF PDU – 1/2

Authentication

Version Type Message length

Checksum Authentication type

SourceAddr

AreaId

0 8 16 31

OSPF common header format

Incorrect routing may cause

large security problems

Make sure LSP is generated by

legitimate routers

Lowest IP address among the

IP addresses assigned to a router

Network

A

Network

B

IP Common Hdr LSA1 … LSAn

OSPF PDU – 2/2

30

Network

A

Network

B

Type 1 LSA

Type 2 LSA

Link-state ID = Advertising router

Smallest IP address

Router ID

IP Common Hdr LSA1 … LSAn

Hdr Body

SNU SCONE lab.

LSA Transmission Directly on IP

– Raw socket

Reliable flooding– Make sure LSA reaches to all routers

• While maintaining the efficiency of forwarding

– Use seqno to detect duplicate

Procedure– Compose a new LSP (Packet) with LSAs periodically

• Increment SEQNOs

– Flood to all links

– When a router receives an LSP

• Check each LSA is new one (SEQNO)

• If new, flood to all interfaces except the one from which the LSP was received

• If not, ignore

Destination IP address?

SNU SCONE lab. 32

Flooding: Example

X

C

A

B D

X

C

A

B D

X

C

A

B

OSPF is a complex protocol

For details, refer to

http://www.csc.kth.se/utbildning/kth/kurser/DD2490

/ipro1-11/lectures/routing_ospf.pdf

46. For the network shown in Figure 3.53, give global distance–vector tables like

those of Tables 3.10 and 3.13 when

(A) Each node knows only the distances to its immediate neighbors.

(B) Each node has reported the information it had in the preceding step to its

immediate neighbors.

(C)Step (b) happens a second time

SNU INC lab. 33

51. A site is shown in Figure 3.55. R1 and R2 are routers; R2 connects to the outside

world. Individual LANs are Ethernets. RB is a bridge-router; it routes traffic

addressed to it and acts as a bridge for other traffic. Subnetting is used inside the

site; ARP is used on each subnet. Unfortunately, host A has been misconfigured and

doesn’t use subnets. Which of B, C, and D can A reach?

SNU INC lab. 34

54. For the network in Figure 3.53, suppose the forwarding tables are all

established as in Exercise 46 and then the C–E link fails. Give:

(A) The tables of A, B, D, and F after C and E have reported the news.

(B) The tables of A and D after their next mutual exchange.

(C) The table of C after A exchanges with it.

SNU INC lab. 35

58. Consider the situation involving the creation of a routing loop in the network of

Figure 3.29 when the A–E link goes down. List all sequences of table updates among

A, B, and C, pertaining to destination E, that lead to the loop. Assume that table

updates are done one at a time, that the split-horizon technique is observed by all

participants, and that A sends its initial report of E’s unreachability to B before C.

You may ignore updates that don’t result in changes.

36

Suppose A notices that link (A, E) is broken① A advertises to B its cost to E is

② C advertises to B that its cost to E is 2

B updates its route to E via C with cost 3

③ A’s link failure message arrives to C

C think E is unreachable

④ B advertises to A that its cost to E is 3

A updates its route to E via B with cost 4

⑤ A advertises to C that its cost to E is 4

59. Suppose a set of routers all use the split-horizon technique; we consider here under

what circumstances it makes a difference if they use poison reverse in addition.

(A) Show that poison reverse makes no difference in the evolution of the routing loop in the

two examples described in Section 3.3.2, given that the hosts involved use split horizon.

(A) Add Poison Reverse to the sequence we have shown in problem 58.

37

Suppose A notices that link (A, E) is broken

① A advertises to B its cost to E is

② C advertises to B that its cost to E is 2

B updates its route to E via C with cost 3

③ A’s link failure message arrives to C

C think E is unreachable

④ B advertises to A that its cost to E is 3

A updates its route to E via B with cost 4

⑤ B also send Poison Reverse to C that its cost to E is ∞ (Poison Reverse) No effect

⑥ A advertises to C that its cost to E is 4

⑦ A also advertise to B that its cost to E is ∞ (Poison Reverse) again with

no effect

(A) Suppose split-horizon routers A and B somehow reach a state in which they forward

traffic for a given destination X toward each other. Describe how this situation will

evolve with and without the use of poison reverse.

(B) Give a sequence of events that leads A and B to a looped state as in (b), even if poison

reverse is used. (Hint: Suppose B and A connect through a very slow link. They each reach

X through a third node, C, and simultaneously advertise their routes to each other.)

(B)

() Similar to (A)

A and B transmit Poison Reverse messages to C

C tra

SNU SCONE lab. 38

A B

Split horizon onlyBoth A and B do not advertise each other

Their route to X will be time out

A B

Split horizon w/ Poison reverseA advertises to B that its cost to X is ∞B advertises to A that its cost to X is ∞Both delete the routes to X immediately

A B

B

B

68. An organization has been assigned the prefix 212.1.1/24 (class C) and wants to

form subnets for four departments, with hosts as follows:

- 75 hosts

- 35 hosts

- 20 hosts

- 18 hosts

There are 148 hosts in all.

(A) Give a possible arrangement of subnet masks to make this possible.

(B) Suggest what the organization might do if department D grows to 32 hosts.

SNU INC lab. 39

72. Table 3.20 is a routing table using CIDR. Address bytes are in hexadecimal. The

notation “/12” in C4.50.0.0/12 denotes a netmask with 12 leading 1 bits: FF.F0.0.0.

Note that the last three entries cover every address and thus serve in lieu of a default

route. State to what next hop the following will be delivered:

(a) C4.5E.13.87

(b) C4.5E.22.09

(c) C3.41.80.02

(d) 5E.43.91.12

(e) C4.6D.31.2E

(f) C4.6B.31.2E

SNU INC lab. 40