SUBNETTING

Historic Network Classes Historically, RFC1700 grouped the unicast ranges into specific sizes called class A, class B, and class C addresses. It also defined class D (multicast) and class E (experimental) addresses, as previously presented.

The unicast address classes A, B, and C defined specifically-sized networks as well as specific address blocks for these networks, as shown in the figure. A company or organization was assigned an entire class A, class B, or class C address block. This use of address space is referred to as classful addressing.

Class A Blocks

A class A address block was designed to support extremely large networks with more than 16 million host addresses. Class A IPv4 addresses used a fixed /8 prefix with the first octet to indicate the network address. The remaining three octets were used for host addresses. To reserve address space for the remaining address classes, all class A addresses required that the most significant bit of the high-order octet be a zero. This meant that there were only 128 possible class A networks, 0.0.0.0 /8 to 127.0.0.0 /8, before taking out the reserved address blocks. Even though the class A addresses reserved one-half of the address space, because of their limit of 128 networks, they could only be allocated to approximately 120 companies or organizations.

Class B Blocks

Class B address space was designed to support the needs of moderate to large size networks with more than 65,000 hosts. A class B IP address used the two high-order octets to indicate the network address. The other two octets specified host addresses. As with class A, address space for the remaining address classes needed to be reserved.

For class B addresses, the most significant two bits of the high-order octet were 10. This restricted the address block for class B to 128.0.0.0 /16 to 191.255.0.0 /16. Class B had slightly more efficient allocation of addresses than class A because it equally divided 25% of the total IPv4 address space among approximately 16,000 networks.

Class C Blocks

The class C address space was the most commonly available of the historic address classes. This address space was intended to provide addresses for small networks with a maximum of 254 hosts.

Class C address blocks used a /24 prefix. This meant that a class C network used only the last octet as host addresses with the three high-order octets used to indicate the network address.

Basic Subnetting

Subnetting allows for creating multiple logical networks from a single address block. Since we use a router to connect these networks together, each interface on a router must have a unique network ID. Every node on that link is on the same network.

We create the subnets by using one or more of the host bits as network bits. This is done by extending the mask to borrow some of the bits from the host portion of the address to create additional network bits. The more host bits used, the more subnets that can be defined. For each bit borrowed, we double the number of subnetworks available. For example, if we borrow 1 bit, we can define 2 subnets. If we

borrow 2 bits, we can have 4 subnets. However, with each bit we borrow, fewer host addresses are available per subnet.

RouterA in the figure has two interfaces to interconnect two networks. Given an address block of 192.168.1.0 /24, we will create two subnets. We borrow one bit from the host portion by using a subnet mask of 255.255.255.128, instead of the original 255.255.255.0 mask. The most significant bit in the last octet is used to distinguish between the two subnets. For one of the subnets, this bit is a "0" and for the other subnet this bit is a "1". Formula for calculating subnets

Use this formula to calculate the number of subnets:

2^n where n = the number of bits borrowed

In this example, the calculation looks like this:

2^1 = 2 subnets

The number of hosts

To calculate the number of hosts per network, we use the formula of 2^n - 2 where n = the number of bits left for hosts.

Applying this formula, (2^7 - 2 = 126) shows that each of these subnets can have 126 hosts.

For each subnet, examine the last octet in binary. The values in these octets for the two networks are:

Subnet 1: 00000000 = 0

Subnet 2: 10000000 = 128

See the figure for the addressing scheme for these networks.

Example with 3 subnets

Next, consider an internetwork that requires three subnets. See the figure.

Again we start with the same 192.168.1.0 /24 address block. Borrowing a single bit would only provide two subnets. To provide more networks, we change the subnet mask to 255.255.255.192 and borrow two bits. This will provide four subnets.

Calculate the subnet with this formula:

2^2 = 4 subnets

The number of hosts

To calculate the number of hosts, begin by examining the last octet. Notice these subnets.

Subnet 0: 0 = 00000000

Subnet 1: 64 = 01000000

Subnet 2: 128 = 10000000

Subnet 3: 192 = 11000000

Apply the host calculation formula.

2^6 - 2 = 62 hosts per subnet

See the figure for the addressing scheme for these networks.

Subnetting a subnet

Subnetting a subnet, or using Variable Length Subnet Mask (VLSM) was designed to maximize addressing efficiency. When identifying the total number of hosts using traditional subnetting, we allocate the same number of addresses for each subnet. If all the subnets have the same requirements for the number hosts, these fixed size address blocks would be efficient. However, most often that is not the case.

For example, the topology in Figure 1 shows a subnet requirement of seven subnets, one for each of the four LANs and one for each of the three WANs. With the given address of 192.168.20.0, we need to borrow 3 bits from the host bits in the last octet to meet our subnet requirement of seven subnets.

These bits are borrowed bits by changing the corresponding subnet mask bits to "1s" to indicate that these bits are now being used as network bits. The last octet of the mask is then represented in binary by 11100000, which is 224. The new mask of 255.255.255.224 is represented with the /27 notation to represent a total of 27 bits for the mask.

In binary this subnet mask is represented as: 11111111.11111111.11111111.11100000

After borrowing three of the host bits to use as network bits, this leaves five host bits. These five bits will allow up to 30 hosts per subnet.

Although we have accomplished the task of dividing the network into an adequate number of networks, it was done with a significant waste of unused addresses. For example, only two addresses are needed in each subnet for the WAN links. There are 28 unused addresses in each of the three WAN subnets that have been locked into address these address blocks. Further, this limits future growth by reducing the total number of subnets available. This inefficient use of addresses is characteristic of classful addressing.

Applying a standard subnetting scheme to scenario is not very efficient and is wasteful. In fact, this example is a good model for showing how subnetting a subnet can be used to maximize address utilization.

Getting More Subnet for Less Hosts

Recall in previous examples we began with the original subnets and gained additional, smaller, subnets to use for the WAN links. Creating smaller each subnet is able to support 2 hosts leaves the original subnets free to be allotted to other devices and prevents many addresses from being wasted.

To create these smaller subnets for the WAN links, begin with 192.168.20.192. We can divide this subnet is to many smaller subnets. To provide address blocks for the WANS with two addresses each, we will borrow three additional host bits to be used as network bits.

Address: 192.168.20.192 In Binary: 11000000.10101000.00010100.11000000 Mask: 255.255.255.252 30 Bits in binary: 11111111.11111111.11111111.11111100

The topology in the figure 2 shows an addressing plan that breaks up the 192.168.20.192 /27 subnets into smaller subnets to provide addresses for the WANs. Doing this reduces the number addresses per subnet to a size appropriate for the WANs. With this addressing, we have subnets 4, 5, and 7 available for future networks, as well as several other subnets available for WANs.