Objectives

After completing this section, students should be able to explain fundamental concepts of computer networking.

IPv4 networking

TCP/IP standards follow a four-layer network model specified in RFC1122.

Figure 11.1: IPv4 addresses and netmasks

IPv4 addresses

An IPv4 address is a 32-bit number, normally expressed in decimal as four octets ranging in value from 0 to 255, separated by dots. The address is divided into two parts: the network part and the host part. All hosts on the same subnet, which can talk to each other directly without a router, have the same network part; the network part identifies the subnet. No two hosts on the same subnet can have the same host part; the host part identifies a particular host on a subnet.

In the modern Internet, the size of an IPv4 subnet is variable. To know which part of an IPv4 address is the network part and which is the host part, an administrator must know the netmask which is assigned to the subnet. The netmask indicates how many bits of the IPv4 address belong to the subnet. The more bits that are available for the host part, the more hosts can be on the subnet.

The lowest possible address on a subnet (host part is all zeros in binary) is sometimes called the network address. The highest possible address on a subnet (host part is all ones in binary) is used for broadcast messages in IPv4, and is called the broadcast address.

Network masks are expressed in two forms. The older syntax for a netmask which uses 24 bits for the network part would read 255.255.255.0. A newer syntax, called CIDR notation, would specify a network prefix of /24. Both forms convey the same information; namely, how many leading bits in the IP address contribute to its network address.

The examples which follow illustrate how the IP address, prefix (netmask), network part, and host part are related.

The special address 127.0.0.1 always points to the local system ("localhost"), and the network 127.0.0.0/8 belongs to the local system, so that it can talk to itself using network protocols.

Figure 11.2: Network routing and DNS concepts

IPv4 routing

Whether using IPv4 or IPv6, network traffic needs to move from host to host and network to network. Each host has a routing table, which tells it how to route traffic for particular networks. The routing table entries will list a destination network, which interface to send the traffic out, and the IP address of any intermediate router that is required to relay the message to its final destination. The routing table entry which matches the destination of the network traffic is used to route it. If two entries match, the one with the longest prefix is used.

If the network traffic does not match a more specific route, the routing table usually has an entry for a default route to the entire IPv4 Internet, 0.0.0.0/0. This default route points to a router on a reachable subnet (that is, on a subnet that has a more specific route in the host's routing table).

If a router receives traffic that is not addressed to it, instead of ignoring it like a normal host, it forwards the traffic based on its own routing table. This may send the traffic directly to the destination host (if the router happens to be on the destination's subnet), or it may be forwarded on to another router. This process of forwarding continues until the traffic reaches its final destination.

In this example, traffic headed for the IP address 192.0.2.102 from this host will be transmitted directly to that destination via the wlo1 wireless interface, because it matches the 192.0.2.0/24 route most closely. Traffic for the IP address 192.168.5.3 will be transmitted directly to that destination via the enp3s0 Ethernet interface, because it matches the 192.168.5.0/24 route most closely.

Traffic to the IP address 10.2.24.1 will be transmitted out the enp3s0 Ethernet interface to a router at 192.168.5.254, which will forward that traffic on to its final destination. That traffic matches the 0.0.0.0/0 route most closely, as there is not a more specific route in the routing table of this host. The router will use its own routing table to determine where to forward that traffic to next.

Names and IP addresses

The IP protocol uses addresses to communicate, but human beings would rather work with names than long and hard-to-remember strings of numbers. DNS, the Domain Name System, is a distributed network of servers that maps host names to IP addresses. In order for name service to work, the host needs to be pointed at a nameserver. This nameserver does not need to be on the same subnet; it just needs to be reachable by the host.

DHCP or static network configuration

Many systems are configured to obtain network settings automatically at boot time. The local configuration files indicate that DHCP should be used and a separate client service queries the network for a server and obtains a lease for network settings.

If a DHCP server is not available, the system must use a static configuration where the network settings are read from a local configuration file. The correct network settings are obtained from the network administrator or architecture team to ensure there are no conflicts with other systems.

Since DHCP uses the hardware address to track assignments, only one address may be assigned per interface with DHCP. Multiple static addresses may be assigned to a single interface. This practice is common in systems hosting services for multiple clients, such as HTTP IP-based hosting. Red Hat Enterprise Linux interfaces typically have an IPv4 address and an IPv6 local-link address, but may have more addresses assigned.

Network interface names

Traditionally, network interfaces in Linux are enumerated as eth0, eth1, eth2, and so on. However, the mechanism which sets these names can cause changes to which interface gets which name as devices are added and removed. The default naming behavior in Red Hat Enterprise Linux 7 is to assign fixed names based on firmware, device topology, and device type. Interface names have the following characters:

For example, the first embedded network interface may be named eno1 and a PCI card network interface may be named enp2s0. The new names make it easier to distinguish the relationship between a port and its name if the user knows both, but the trade off is that users cannot assume a system with one interface calls that interface eth0.