����ý

In networking, layering means to break up the sending of messages into separate components and activities. Each component handles a different part of the communication. This can be referred to as a protocol stack. The Transmission Control Protocol/Internet Protocol (TCP/IP) is a protocol stack used when connecting a device to the internet.

There are four layers to be considered:

• Application layer - encodes/decodes the message into a form that is understood by the sender and the recipient devices using protocols like HTTP, file transfer protocol (FTP) and SMTP.
• Transport layer - breaks down the message into small pieces called packets. Each packet is given a packet number and the total number of packets. The recipient uses this information to assemble the packets in the correct order. It also allows the recipient to see if there are any missing packets.
• Internet layer - adds the sender’s IP address and that of the recipient. The network then knows where to send the message, and where it came from. This is sometimes called the network layer.
• Link layer - enables the physical transfer of packets between nodes on a network, and between one network and another.

Layering enables standards to be developed, but also to be adapted to new hardware and software over time. For example, different software packages (applications) may use the same transport, internet and link layers but may each use their own application layer. The way the program encodes the message changes - the rest of the communication method remains the same.

IP addressing

At the internet layer, the IP - Internet Protocol - part of the TCP/IP protocol stack needs each node to have an IP address assigned to it. This address consists of four sets of up to three digits, each with a maximum value of 255. Each set is separated by dots. An example is:

192.168.0.254

When a node wants to send a message to another node, it uses the recipient node's IP address as the destination.

The method of addressing using four sets of digits held in 32 bits is called IPv4. When IPv4 was devised, the range of addresses it allowed was deemed to be sufficient. However, the huge increase in internet-enabled devices led to IPv4 running out of addresses. A new version - IPv6 - has been introduced, significantly increasing the number of possible addresses to trillions. IPv6 addresses are held in 128 bits.

Transmissions over a network can be extremely large. To send a large message in one go can be impractical, as both the sending and receiving node are tied up with one communication.

Packet switching

TCP/IP uses packet switching to send data across a network. In packet switching, messages are broken up into very small pieces, called packets. Each packet consists of two parts:

• header - this includes the sender's and recipient's IP addresses, the packet number, the total number of packets the message contains, plus the details of any protocols used
• payload - this is part of the message itself

The packets are sent individually across the network and put back together to reform the message at the other end.

Packets may or may not follow the same route - they are sent along whichever route gives the quickest transmission. Therefore, packets may arrive out of sequence.

The recipient takes each packet, makes a note of its number and then puts it into the correct order. A packet switched network can therefore be fault tolerant. If a particular connection fails, the packets can be sent via a different route and then reassembled in the correct order. This is possible even if some packets are delayed.