[3 pages — 2378 words]


INTRODUCTION

IP resides at the network layer of the OSI communications model and provides the basic unit of data transfer, which includes addressing, routing, and fragmentation. The transport layer of this same model resides above the network layer and provides station-to-station communication and a common interface to the application layer. This implies reliable communication, which is either accomplished at the transport layer or at the application layer. With control networks, this is usually accomplished at the application layer since many control networks were designed before the popularity of TCP/IP took hold. Still, there are some control network protocols, such as MODBUS/TCP, which do rely upon the guaranteed delivery mechanism of TCP and there may be more in the future. Actually, at the transport layer of the TCP/IP stack there are two transport protocols, each of which find service in control networks. The User Datagram Protocol (UDP) and the Transmission Control Protocol (TCP) will both be discussed in this lesson.


USER DATAGRAM PROTOCOL

UDP provides a connectionless and unreliable transport service since it does not issue acknowledgments to the sender upon receipt of data nor does it inform the sender that data was lost. Data integrity can suffer by dropped packets, mis-sequenced packets or by the receipt of duplicate packets. Any of these situations can occur without the knowledge of the sender. It appears that UDP is no better than the IP protocol but there is one big difference. UDP introduces the concept of port numbers, which are used by the application layer that resides above UDP. Port numbers have significance in terms of actions requested by the application itself that require a particular response by the receiving station.

The UDP header is short and simple. Only eight bytes are required in the header. Source and destination ports are each 16-bits long and, therefore, require four bytes. The message length of 16-bits indicates the length of the header and attached data. A 16-bit checksum is used to check the validity of the header and data. The UDP header and attached data, which comes from the application layer, are encapsulated into the IP data field. An IP header, which provides station addressing, precedes the UDP datagram and the complete IP datagram is encapsulated into the frame of the data link layer technology used, such as Ethernet, and sent to the desired station where the complete process is reversed. Notice that the only contribution UDP provided was the assignment of port numbers for use by the application layer. If UDP is to be used, the application layer must worry about acknowledging message receipt, correctly ordering received packets into meaningful messages, discarding duplicate packets and requesting retransmission of faulty packets since UDP does not provide this service. However, if the application layer was originally designed to provide this reliability of service, there is no reason to have the transport layer duplicate effort so UDP makes sense. UDP has low overhead and executes quickly, making it attractive for control networks.

Figure 1 — The UDP header is quite short and, along with its data, it is encapsulated into an IP datagram.

PORT NUMBERS

UDP introduces the port number concept. When a station receives an UDP datagram, it serves up the port number to the application layer, which then allocates a buffer area for the attached data. The port number has significance since it identifies a particular application. Since there are many port number possibilities, several different applications can be simultaneously supported on the same station.

Port numbers are 16-bits long and are classified as being assigned, registered or dynamic. Assigned port numbers in the range of zero to 1023 have been defined by the Internet Assigned Numbers Authority (IANA) for various applications that are considered part of the TCP/IP protocol suite. These applications include TELNET, FTP, and other popular Internet applications. These port numbers are termed “well known” and cannot be used by other applications. The remainder of the port assignments is classified as being either registered or dynamic. Registered means that an organization wants to define some level of functionality and has registered a unique port number with the IANA. Other organizations are to respect this assignment and not use either assigned or registered port numbers. Finally, the dynamic port numbers are used in a random manner by a requesting station to signify the source port of an application request.

For example, if station A requests a trivial file transfer protocol (TFTP) service (TFTP happens to use UDP) from station B, it would insert a 69 (a well-known port assignment indicating TFTP services) into its destination port. A dynamic (random but non-conflicting) number will be put in its source port and the request will be sent to station B. Station B would receive the request and recognize it as a TFTP request from station A since port number 69 was used. Station B would then begin executing the process. But it would insert a 69 in its source port, and the dynamic number that was generated by station A in its destination port, and send the response to station A. Station B knows how to handle this particular application since it recognizes both the source and destination port numbers. The use of port numbers along with the IP address creates what is called a socket, which can be represented as <netid, hostid, portid>. As long as there is structure to the assignment of ports, the socket assignment becomes a unique representation of a particular application on the complete IP network.

PORT NUMBER ASSIGNMENTS

Both UDP and TCP use port numbers and, if possible, the same assignment is used for both. The 16-bit numbers are classified as either assigned (well-known), registered or dynamic (private). Port number assignments are maintained by the Internet Assigned Numbers Authority (IANA) and the complete list can be found at http://www.iana.org/assignments/port-numbers under port numbers. Numbers in the range of zero to 1023 are classified as well-known and are used by common processes. However, individual organizations can register numbers in the range of 1024 to 49151 for proprietary purposes and will not be used by other organizations. Looking at the list, we find some companies in our industry. Opto22 has registered 22000 and 22001 for their SNAP I/O and Opto Control products. The BACnet building automation protocol has registered 47808. DeviceNet and ControlNet intend to use 44818. Interestingly, the MODBUS/TCP specification calls for port 502, which is in the well-known port group. Port numbers 49152 to 65535 are considered as either private or dynamic and can be used by anyone.

(No part of this article may be reproduced without the written consent of the Industrial Ethernet University.)