This lecture is divided into hyperlinked sections

Introduction
Why do we need a Protocol Architecture?
What is a protocol?
The TCP/ IP Protocol Architecture Structure
Operation of TCP/ IP
TCP/ IP Applications
TCP/ IP Dependent Protocols
Conclusion
Tutorial Questions

We shall examine computer communications and discuss the requirements and then compare the OSI model with the Internet model, the TCP/ IP suite.

We will discover what protocols are and why they are needed and look at several protocols and observe their function.

We shall then see how TCP/IP delivers a message across a network and see some applications that use TCP/IP.

We will see the use of port numbering to identify processes across an internetwork

Why the need for Protocol Architecture?

When any two computer based entities wish to exchange data there must exist more than just a physical data path between the two communicating devices. There has to be some amount of formal co-operation between the two devices to enable the data transfer to take place. Some typical tasks that need to be performed for the data transfer to take place are:

1. The source system (client) must set up the communication path or inform a remote system it wishes to communicate with it
   
2. The source system (client) must determine whether the remote system is ready to accept data
3. The data transfer application on the source system must make sure that the file management program (daemon) within the remote system is ready to accept the data from the source
4. If there is any incompatibility between the two system's file formats, one of the systems must perform a format translation e.g. big endian to little endian
   
5. If data are lost, there must be some recovery system
6. When the data transfer is complete, the systems must inform each other of their readiness to break the connection

In a protocol architecture, the logic modules are arranged in a vertical stack. Each layer of this stack performs a subset of the entire logic necessary to communicate with a remote system. Layers should be designed so that changes in one layer do not affect any higher or lower layers.

This stack of logic modules must be implemented in both of the machines that wish to communicate. Communication is achieved by corresponding, or peer levels in both machines communicating with each other. Each of these peer levels communicates according to a set of rules or conventions that are known as a protocol.

What is a protocol

A protocol is the formal code of behaviour for a system, or to be more precise the set form in which data must be presented for handling by a particular computer configuration. A protocol defines the set of rules governing the exchange of data between two entities such as user application programs, file transfer packages, email etc.

The key features of a protocol are:

• Syntax. This is concerned with the format of the data blocks
• Semantics. This includes control data for co-ordination and error handling
• Timing. This allows for speed matching and sequencing.

The TCP/ IP Protocol Architecture Structure

The TCP/ IP suite has come to dominate computer communications. The OSI protocols were still in development when the TCP/ IP protocols had already been tested and implemented. When businesses realised the need for interoperability, it was only TCP/ IP that was ready to be used. Another factor is that the OSI model is too complicated and needs seven layers to do what TCP/ IP does in five.

For reasons best known to themselves the Internet community have chosen to divide the functionality of the protocol stack in a different manner to OSI but not without maintaining the compatibility. The layers correspond roughly as shown below.

Operation of TCP/ IP

To understand how one application that resides on one computer operates, let us consider the case of a PC on an Ethernet network that wishes to communicate with another machine on a different network. For our purposes two separate Ethernet networks separated by a router will help illustrate the process. For successful communication to take place, every entity in the overall system must have a unique address. The host is identified with an IP address so IP is implemented in all end systems and routers.

Compare this with surface post. It is necessary to have the address of the house to which you want your letter to be delivered, and this is written on the envelope. This can be likened to the IP address. However a PC may have many applications running in seprate processes (identified by port numbers) that are able to communicate and these may be compared with the inhabitants of the house to which you send the letter. When you write a letter, you specify the person within the address to which the letter is bound. The post office sorting system only looks at the postcode and address written on the envelope and delivers it to the house (IP address). Once the letter has arrived at the house, it is further sorted by the person who picks up the letters and passed to the person whose name precedes the address (port number).

This is the same as computer addressing where the host has an IP address and the application process within the host has an identity too. The address for the application process within the host is known as a port. It is TCP at the Transport layer that looks at the port address and delivers any incoming data to the proper application process specified by the port number.

Let us consider the operation where a process running on one host wishes to send a message to another process on a remote host.

To help with the following discussion, please refer to the diagram above. The process (application) is associated with an arbitrary port e.g. 2343 on host A and wishes to communicate with a process associated with port 80 on host B. The process on host A sends the message to TCP with instructions to send to host B, port 80. TCP hands the message down to IP with instructions to send to host B. IP does not need to know the destination port as this has been taken care of by TCP. IP then passes the message down to the network access layer along with the IP address of host B (193.60.61.124), Ethernet in this case with instructions to send the message to a router R which is the first hop (default gateway) on the route to host B.

For this operation to take place, extra information other than the data generated by the application has to be transmitted too, see the figure below. For each piece of data passed to TCP, there is some control information too that is added by TCP in the form of a header, a few bytes of information that is prepended to the data. This forms a TCP segment. This information will be used when the TCP segment is examined by host B to pass the data to the correct port. It is not used or seen at any other point between host A and B.

Some of the information included in the TCP header are:

• Destination port number for the remote host's application - 80 in this case
• Source port - arbitrary on the host PC
  
• Sequence number. TCP numbers the segments in case they arrive at the destination out of order so that TCP at host B can re-order the segments correctly
• Checksum. This is in the form of a cyclic redundancy check (CRC) and is used to check that the data have arrived correctly and not been corrupted during the transmission process.

This IP datagram is presented to the network access layer for transmission across the first subnetwork in its journey towards its destination. The network access layer then appends its own header containing information that is required to transfer data across this first subnetwork.

Information contained in the network access layer includes:

• Subnetwork destination address. Here this will be the intermediate router .
• Facilities request. The network access protocol may request use of subnetwork facilities such as priority.

The data will then be delivered to host B where firstly the network header is stripped off. The IP header is removed too, revealing the TCP data (port address). The TCP layer in host B then is able to remove the TCP control information and pass the (original) data to the correct port for the residing application process to decode and present to the user.

TCP/IP applications

Several applications have been written to operate within the TCP/IP environment. The three most common are:

• SMTP. This is simple mail transfer protocol and provides for email. SMTP does not specify the application used to create the email message, it merely accepts the message and makes use of TCP to send the message to the TCP entity residing on a destination machine.
• FTP. File transfer protocol is used to send files from one machine to another under user command.
• TELNET. This provides a remote logon facility to enable a user to log into a remote machine as if directly connected to that computer. Terminal traffic is carried between the user and the TELNET server by TCP.

The TCP/IP protocol suites occupy two levels in the system model, those of Transport and Internet layers respectively.

The Internet layer, i.e. that which deals with addressing using IP numbers, is a connectionless mode while the transport layer is connection oriented if Transmission Control Protocol (TCP) is used but again connectionless if User Datagram Protocol (UDP) is used.

Connectionless mode is when the end entity is not contacted before the data transfer and no defined data path is required to exist.

Connection oriented mode is where a connection must be established with the end entity before data transfer may commence.

There are situations where either are the best choice for the application in use. The most common Internet applications are :
 

• SMTP, the Simple Mail Transfer Protocol. A simple Store-and-Forward service for text information.
• HTTP, Hypertext Transfer Protocol which is used for request and delivery of web pages.
  
• FTP, the File Transfer Protocol, which provides for the transfer of files using 'get' and 'put'.
• TELNET a virtual terminal service providing remote login facilities to any remote computer supporting TELNET.
• DNS the Domain Name System. Hosts are frequently known by alphabetical names but to access them an IP number is needed. DNS provides the translation between names and IP addresses in a client/ server type of relationship.
• SNMP the Simple Network Management Protocol. This allows for monitoring and management of network resources remotely Note that SNMP is not confined to TCP/ IP networks.

The relationship between the protocols and the end-to-end services is shown below.
 

Figure: Protocol layer relationships

In the figure above it can be seen that the Internet layer provides a delivery service independent of the underlying technology of the interface layer. To do this the Internet layer requires global addressing facilities and to be able to route datagrams to the required destination. The selected path may traverse a number of different networks with differing interface layers. This latter requirement may affect the operation of the Internet layer.

There is one protocol not previously mentioned, that of ICMP Internet Control Message Protocol. This is inside the IP layer and provides some simple control facilities independent of SNMP.

Conclusion

We have seen that a protocol architecture is required to allow for a structured set of rules to be established that will allow communications to take place between a client and the server.

Within the communication architecture reside protocols which are sets of rules for each individual operation to take place.

We have seen that a practical protocol architecture in use is TCP/ IP, how it fits in with the OSI model and how it is used in a communication arena.

We named several of the most popular applications that use TCP/ IP, namely HTTP, SMTP, FTP and TELNET.

The transport layer makes use of port numbers to identify processes on the end systems.

The Internet layer makes use of IP to address individual end systems.

Tutorial questions

• Which layer of the TCP/ IP model contains IP addressing information?
  
• Which layer of the TCP/ IP model contains application process identification details?
• Why would the host (client) use an arbitrary port number in communications but the server use a specified port number e.g. 80 for HTTP?
  
• Why is it necessary to include the host (client) IP address when requesting a service from the server?
  
• Why is it necessary to include the host (client) port address when requesting a service from the server?

In May of 1974, Vint Cerf and Bob Kahn of ARPA began to develop methods and protocols for internetworking. This is the ability to communicate across different and arbitrary packet switched networks. The proposal in their paper was enthusiastically received and their proposal was refined and with contributions from ARPANET and Cyclades (a French network project) and other projects from around the world formed the basis of what was eventually to become Transmission Control Protocol (TCP) and Internet Protocol (IP). These in turn became the foundations for the TCP/IP protocol suite. In 1980, experimentation with TCP/IP commenced.

This provided the underpinning of what was to become the Internet as we know it today with Arpanet being just one of a collection of interconnected networks.

Between 1982 and 1983, Arpanet converted form its original NCP (Network Control Protocol) to TCP/IP. Many networks throughout the world were interconnected using this technology.

Worldwide realisation of the usefulness of the technology of networking brought the National Science Foundation (NSF) to give support to other computer science research groups. In 1986, NSF extended support for all disciplines of the general research community with the NSFNET backbone. Eventually NSFNET offered interconnection via its backbone to regional packet switched networks across the United States. The Joint Academic Network JANET was formed around this time to interconnect European universities and colleges.

In 1990 the ARPANET was shut down.

---

(c)MMClements 2008                                                      Back to top of Page