IPv4 addressing and classful addressing

Why it exists

Every interface on the Internet needs a unique number so that packets can be routed to it. IPv4 uses a $32$-bit number. A routing decision cannot look up each of billions of hosts, so the number is split into a network part (common to all machines on one network, used by routers) and a host part (identifies the machine within the network). The original rule for the split was fixed by the first bits of the address: the classes. The classes are obsolete in practice (replaced by CIDR, next page) but their counts and ranges are still asked in GATE.

Prerequisites

You need the network layer's job: end-to-end delivery by address (layering).

Precise definitions

An IPv4 address is a $32$-bit number, written in dotted decimal: four bytes in decimal separated by dots. Example: 172.16.37.5 is 10101100.00010000.00100101.00000101. There are $2^{32} = 4\,294\,967\,296$ addresses.

An address identifies an interface, not a host: a router with three network cards has three addresses, one in each network it connects, and a laptop with Ethernet and Wi-Fi has two.

Classful addressing assigned the split from the leading bits of the address:

• Class A: leading bit $0$; first octet $1$ to $126$ (usable); $8$ network bits, $24$ host bits.
• Class B: leading bits $10$; first octet $128$ to $191$; $16 + 16$.
• Class C: leading bits $110$; first octet $192$ to $223$; $24 + 8$.
• Class D: leading bits $1110$; first octet $224$ to $239$: multicast (group addresses, no network/host split).
• Class E: leading bits $1111$; first octet $240$ to $255$: reserved.

The number of networks is $2^{\text{network bits} - \text{leading bits}}$; for class A it is $2^7 = 128$ minus the two reserved first octets $0$ and $127$, so $126$. The number of hosts per network is $2^{\text{host bits}} - 2$: the host part all zeros names the network itself, the host part all ones is the broadcast address. A class A network has $16\,777\,214$ hosts, class B $65\,534$, class C $254$.

A subnet mask has ones in the network (and subnet) positions and zeros in the host positions; a bitwise AND of an address and the mask gives the network address. Default masks: class A 255.0.0.0, class B 255.255.0.0, class C 255.255.255.0.

Special addresses (quoted from the RFCs; the exam usually needs only the first four):

How it works

Reading an address

To find the class, look at the first octet in binary (or the decimal ranges above). To find the network address, keep the network bits and set the host bits to zero. Example: 200.100.50.12: first octet $200$ lies in $192 \dots 223$, class C, network part 200.100.50, network address 200.100.50.0, broadcast address 200.100.50.255, $254$ hosts.

A host decides whether a destination is on its own network by comparing network parts: if the destination's network part equals its own, the packet is sent directly (the destination MAC address is found with ARP); otherwise it is sent to the default router. A router uses the network part to look up the route (routing).

Why classes were replaced

A company with $2000$ computers did not fit in a class C ($254$ hosts) and took a class B ($65\,534$ hosts), wasting about $63\,000$ addresses. Class B addresses ran out, class C tables grew huge. Subnetting let an organisation split its network internally; CIDR removed the classes altogether (CIDR and subnetting).

Worked examples

Cheat box

Traps and GATE patterns

Connections

CIDR keeps the idea of a network part and a host part but lets the split fall at any bit (CIDR and subnetting). Private addresses make NAT necessary (NAT). Hosts find the MAC address of a neighbour with ARP (ARP, ICMP and DHCP), and the IP header is described on IPv4 fragmentation.

PYQs

If an IP network uses a subnet mask of
255.255.240.0, the maximum number of IP
addresses that can be assigned to network interfaces
is ______. (answer in integer)

Consider an enterprise network with two Ethernet
segments, a web server and a firewall, connected via

COMPUTER NETWORK
three routers as shown below:

What is the number of subnets inside the enterprise
network?

Consider three machines M, N, and P with IP
addresses 100.10.5.2, 100.10.5.5, and 100.10.5.6
respectively.
The
subnet
mask
is
set
to
255.255.255.252 for all the three machines. Which
one of the following is true?

In the network 200.10.11.144/27, the fourth octet
(in decimal) of the last IP address of the network
which can be assigned to a host is ______.

Every host in an IPv4 network has a 1-second
resolution real-time clock with battery backup. Each
host needs to generate up to 1000 unique identifiers
per second. Assume that each host has a globally
unique IPv4 address. Design a 50-bit globally unique
ID for this purpose. After what period (in seconds)
will the identifiers generated by a host wrap around?

In the IPv4 addressing format, the number of
networks allowed under Class C addresses is

Suppose computers A and B have IP addresses
10.105.1.113 and 10.105.1.91 respectively. They
both use the same netmask N. Which of the values
of N given below should not be used if A and B
should belong to the same network?

If a class B network on the Internet has a subnet mask
of 255.255.248.0, what is the maximum number of
hosts per subnet?

The address of a class B host is to be split into
subnets with a 6-bit subnet number. What is the
maximum number of subnets and the maximum
number of hosts in each subnet?

Two computers C1 and C2 are configured as follows:
C1 → IP: 203.197.2.53, Mask: 255.255.128.0
C2 → IP: 203.197.75.201, Mask: 255.255.192.0

Which statement is true?

An organization has a class B network and wishes to
form subnets for 64 departments. The subnet mask
would be:

Self-check

Revision card

References