The IP protocol identifies each computer connected to the network by its corresponding address. This address is a 32 bits number in Ipv4, that has to be unique for each server or computer, which we will call “host”. The IP addresses are usually represented as four decimal ciphers, of 8 bits each, separated by dots.

 The Internet address (IP Address) is used to identify both the host and the network to which it belongs, in a way that distinguishing the hosts connected to the same network, is possible. With this purpose, and considering that there are different sizes networks connected in Internet, five kinds of different addresses have been settled.

 The idea is that according the network's topology, more bits are used for the network’s address than for the host’s address, or viceversa. Figure I-8 shows the three kinds of addresses, A,B and C, to which we added type D and E to represent all the receptors (“multicast”) and for future purpose.

Figure I. IP address’s kind schema.

 The first bits (shaded) define the kind of address that the next bits carry. As a reference, with 8 bits (each part between dots), you can have 256 different values (from 00000000=0 to 11111111=255). For example, an IP address would be 192.228.17.57 ( in bits 11000000 11100100 00010001 00111001), representing a C kind address.

 Let us see the most important characteristics of each kind:

-       Class A: Are those that in their first 8 bits have a fixed bit 0 (indicating class A) and 7 variable bits. That is, they can differentiate from 00000000=0 to 0111111=127. Nevertheless, the 0 and the 127 are reserved, so there might be only126 potential addresses class A that correspond to the first byte of the address (the first number before the dot). The other three bytes (24 bites) are available for each one of the hosts belonging to the same network. This means that there might exist 2²⁴ = 16.387.064 computers or servers in each network of this kind. This kind of addresses is used in very extensive networks, but we must keep in mind that there can be only 126 networks of this size. That is why big commercial networks use them, although that there are few organizations that get a “class A” address. Generally big organizations use one or more “class B” networks.

-       Class B: These addresses use in their first byte the 10 bits fixed (indicating class B) and with the rest of the first byte’s bits admits addresses from 128=10000000 and 191=1011111, including both. In this case, the network’s identifier is obtained from the first two bytes (16 bits – 2 bits already used = 14 bits) of the address, having to be a value between 128.1 and 191.254 (it is not possible to use values 0 and 255 because they have a special meaning). Therefore, there will be 2¹⁴ = 16.384 different networks class B. The last two bytes of the address constitute the host’s identifier allowing a maximum of 2¹⁶ = 64.516 computers in the same network. Addresses of this kind should be enough for most of the big organizations. In case the number of computers needed is bigger, it would be possible to have more than one “class B” address, avoiding so the use of a “class A” one.

-       Class C: In this case the value of the first byte will include the 110 bits fixed (indicating class C) and then the first byte must be between 192=11000000 and 223=11011111, including both values. This third kind of address use the three first bytes for the network’s number, with a range that goes from  192.1.1 up to 223.254.254. Therefore, there will be 2²¹ = 2.097.152 different networks class C. So, we have one byte (8 bits) free for the host, allowing a maximum of 254 computers connected to each network, since 2⁸ =256 but the 0 and the 255 are not used.

-       Class D: The addresses that begin with the 1110 bits fixed followed with the “multicast” address that is for every destiny, are usually called class D.

-       Class E: We usually call class E the addresses kept for future use.

 Table I resumes the important characteristics of class A, B and C addresses.

Internet’s IP addresses table.
Class	First byte	Network’s identification	Host’s identification	Networks number	Hosts number
A	1 .. 126	1 byte	3 byte	126	16.387.064
B	128 .. 191	2 byte	2 byte	16.256	64.516
C	192 .. 223	3 byte	1 byte	2.064.512	254

Table I. IP addresses classes.

 It is important to notice that the values 0 and 255 in any address’s byte can not be used normally because they have other specific tasks.

 The number 0 is set apart for machines that do not know their address. It can be used either in the network’s identification in machines that still not know the network’s number to which they are connected. Or in the host’s identification for machines that do not know their host number in the network; or in both situations.

 As we have seen, number 255, is reserved for the multicast. This is necessary when we want a message to be visible by every system connected to the same network. This might be useful if we need to send the same packet to a specific number of systems, resulting more efficient than sending the wanted information individually to each one. Another situation for the use of multicast is when you want to change the computer’s domain name for its corresponding IP number and you do not know the closest address server’s name

Usually, when the use of multicast is required, we use an address made of the ordinary network’s identifier and the number 255 (all ones in binary) in each byte that identifies the host. Nevertheless, due to convenience it is also allowed the use of the number 255.255.255.255 for the same purpose, so it is easier to refer to every network’s systems.

Subnets

 In some extensive organizations it might arise the need to divide the network into smaller ones (subnets). Then, bits fixed for the host’s addresses in each class, are divided into two groups. Part of them defines the subnet, and the rest the computer inside the subnet. It is necessary to consider that for the external world (that handles the network’s addresses only) this decision of the local network, or computer’s group, is not relevant. Then, each subnet can manage its host’s address bits as it wishes.

The division of the host address bits into subnets is done through a “mask”, which is a well defined bits standard that determines which bits to use in the network’s identification, and which ones identify the computer within the subnet.

 For example, let us consider the following values for a class C address:

IP address:        192.228.15.57      In bits:            1100000.11100100.00010001.00111001

Mask:               255.255.255.22    In bits:            1111111.11111111.11111111.11100000

Logic AND :       192.228.17.32      In bits:            1100000.11100100.00010001.00100000

 That is to say, that from the last 8 bits, for the host’s address in class C, the first 3 bits (in blue) determine the subnet; and the last 5 (in yellow) masks (bring to 0) the computer’s address within the subnet. The binary value of the original last 5 bits will determine the computer’s address within the subnet. In this example, the address 192.228.15.57 determines the class C network 192.228.15, and inside this one the subnet number 1= 001 and in the subnet 1, the computer 25=11001.

 Figure 2 shows an example with 3 subnets, LAN X, LAN Y and LAN Z, following up the class C case seen above. The set is seen for the rest of Internet as the address 192.228.17.x, where x represents the class C host’s address.

Figure 2. Subnets.

 ROUTES DETERMINATION

 It is clear that an important part of the protocol's performance depends on the way the routes that the packets will follow, are determined. We use algorithms that work with different sorts of values to determine the route, and we call them “metrics”. These metrics, either automatically calculated or defined by the network’s manager, are used to make and adapt the tables used by the routers and intermediate nodes. The most common metrics are:

-         Path’s Length: It is the most used metric. Sometimes, costs are assigned to each way’s section, then the path’s length is the sum of each jump’s cost. In other cases the length is calculated as the sum of the jumps the packet has to undertake.

-         Reliability: It is an arbitrary value assigned by the network’s manager. In some case it reveals the Bits Error Rate of the section. In other cases is the probability that the link fall down, the recovery ability or a combination of both.

-         Delay: Represents the time needed to transport a packet from one point to another. This value depends on many factors: section’s bandwidth, distance to be covered, queues made in the routers, link’s congestion, etc. This metric is generally one of the most useful.

-         Bandwidth: It is measure of the information’s transmission speed expected in a link. Generally, the higher the better, even though all by itself it is not a definitive metric. For example, we have to take into account factors like the queues in the routers, etc, which make that the link’s speed is wasted.

-         Load: Indicates how busy an element in the network, (like a router), is. For example, the processor’s usage percentage is a factor that influences this metric, determining the process capability in packets per second.

-         Cost: It refers to the monetary cost. In some cases it is better to route a packet by a longer but cheaper route.

Roberto Rossi
Universidad Blas Pascal