The emergence of packet switching & packet broadcasting networks is because of the requirement to allow the computer user to be able to access resources which are not available on a single system. Similarly, the resources of a single network are often insufficient to meet users' requirements. The network that might be of interest may exhibit several differences; therefore it would be impractical to consider merging them into a single network. More appropriately, the ability to interconnect various networks is necessary. This enables communications to occur between two stations, each located on either network. This interconnection of networks is referred to as INTERNETWORKING.

From the user's point of view, an interconnected set of networks may appear as a larger network. However, if each of the constituent networks retains its identity, & special mechanisms are required for communicating across multiple networks, then the entire configuration is often called an internet & each of the constituent networks are subnetworks.[1]

Asynchronous Transfer Mode (ATM) is form of data transmission which permits voice, video & data to be sent along the same network. In the past, separate networks were used to transfer each of these: voice over the phone, video over cable networks and data over an internetwork.

ATM is a cell-based, connection-oriented, switching & multiplexing technology designed to be a fast, general-purpose transfer mode for multiple services. It allows a virtually unlimited number of users to have dedicated high-speed (upto gigabits per second) connections with each other & high performance network servers. ATM is a technology defined by protocol standards created by the ITU-T, ANSI, ETSI & the ATM Forum.

ATM is also intended to be used across LANs & WANs to provide seamless connectivity therefore making it a suitable technology for internetworking.

ATM is an acronym which stands for Asynchronous Transfer Mode. This technology inherently has aspects such as transport, switching, network management & customer services. ATM could be referred to as a cell relay technology because traffic is carried in small, fixed-length packets called cells. Each cell (Fig 2.1a) is 53 bytes in length, 5 bytes of which is a header & 48 bytes reserved for the payload. This fixed length cell provides advantages from previous packet technology. Since these are short cells, they can be switched in hardware therefore ATM can be switched rapidly & economically. Another advantage is that queuing delays caused by long, variable-length frames can be reduced to the wait time for a 53 byte cell. This allows time dependent video & voice to be transported.

The two main transport technologies available today are Time Domain Multiplexing, which is used by the traditional digital telephony & packet switching systems such as X.25.

Time Domain Multiplexing systems allow fixed bandwidth channels (Fig 2.1b). This is ideal for continuous bit rate signals such as voice & video. But for data transmission (i.e. bursty traffic) this is wasteful of network resources. Another point is that for example in the U.S. there is no standard transport between 1.5Mbps & 45Mbps.

Packet transfer technologies (Fig 2.1c) allow for statistical multiplexing of frames onto the transport medium. This is ideal for variable bit rate data transport, such as computer data. However, for applications such as voice & video it can cause a relatively high fixed delay as well as unpredictable queuing delays.

ATM can provide both, continuous bit rate (CBR) and variable bit rate (VBR) transports:

ATM Cell Relay is a packet switching technology that can provide high quality CBR service and economical VBR transport. In Fig 2.1d, CBR service is provided by allocating equally spaced slots. Data cells can fill up the remaining bandwidth.

ATM is a connection oriented transfer mode. The connection between points must be made prior to the transfer. The path between the stations must also be established. In this form of transmission, since an established connection already exists, a confirmation from the receiver is not required.

ATM is also capable of supporting connectionless traffic, such as IP (Internet Protocol).

With reference to Fig 2.2a:

Within an internet, devices used to support end user applications or services are called End Systems. (e.g. ES A & ES B are connected to Subnetwork A & B respectively.)

The Intermediate System (e.g. IS) which could be a bridge or a router allow the subnetworks to connect to each other. Bridges are used when the subnetworks are operating with the same protocol. Else a router is used.

A bridge operates at layer 2 of the OSI 7-layer model. It does not modify the content of the packets nor does it add anything to the packet. It acts as an address filter by picking up packets from one subnetwork that are intended for a destination on another one, and passing these on. A router operates at layer 3 & routes packets between potentially different networks. It uses an internet protocol common to all routers and hosts of the network. The most important architectural approach to internetworking is the connectionless router [1].

The most widely used internetworking protocol, simply called IP (Internet Protocol) is discussed in the next section.

TCP/IP (Transmission Control Protocol / Internet Protocol) are one of the most extensively used protocol suites in the area of computer communications. The Internet & especially the World Wide Web (WWW) which is also based on IP are gaining popularity at an exponential rate. The increasing number of subscribers evidences this. Therefore, ATM networks must also be capable of supporting these protocols. In ATM networks, TCP/IP protocols will be placed on top of ATM. Fig 2.3a shows the position of the protocols with respect to each other: [1]

IP does not provide error correction, Quality of Service (QoS) guarantees nor information regarding successful delivery about its packets. In case the destination address is unknown, the Internet Control Message Protocol (ICMP) can be used to inform the sender. ICMP messages are encapsulated & conveyed in IP packets. IP data packets need not arrive in the same order they were transmitted in. As a result higher layer protocols such as TCP are responsible for ensuring data integrity. TCP is a connection oriented reliable transport protocol. UDP (User Datagram Protocol) is a connectionless transport protocol which is unreliable.

The principles of carrying IP traffic over an ATM based network are discussed in the following RFCs. These are based on the Internet Engineering Task Force (IETF) specifications

• RFC 1483 'Multiprotocol encapsulation over ATM adaptation layer 5' (IETF, 1993)
• RFC 1577 'Classical IP & ARP over ATM' (IETF, 1994)
• RFC 1626 'Default IP MPU for use over ATM AAL5' (IETF, 1994)
• RFC 1755 'ATM signaling support for IP over ATM' (IETF, 1995)
• RFC 2020 'Support for multicast over UNI 3.0/3.1-based ATM networks' (IETF, 1996) and several internet drafts such as "Next hop resolution protocol" (NHRP) (IETF, 1996) , and 'IP broadcast over ATM' (IETF, 1996b)

[2]

The Internet suite of protocols has become the de facto standard for open systems. Consequently, many legacy systems use the IP as their network layer protocol. IP over ATM is a service that provides a mechanism by which IP can be transported over the ATM network [3].

The two main processes specified by IP over ATM are Packet Encapsulation & Address Resolution.

 2.3.1 Packet Encapsulation

At the network layer, this mechanism permits multiple packet types to be multiplexed on the same connection. The main advantage of this is that it allows connection reuse, therefore providing a reduction in connection setup time. It works by appending a multiplexing field that enables the receiving node to identify the packet type. The encapsulated packet is carried over the ATM AAL5.

AAL5 is one of four ATM Adaptation Layers (AALs). AAL5 supports connection oriented variable bit rate data services. It is relatively small when compared to AAL3/4 (which is classed as one layer). But AAL5 does not have error recovery or built in data transmission. This tradeoff provides a smaller bandwidth overhead, simpler processing needs and reduces the implementation complexity.

RFC 1483 [4] discusses two methods of packet encapsulation:

• LLC/SNAP encapsulation
• VC multiplexing

When several protocols are carried over the same VC (Virtual Channel - a communication channel that provides for sequential & unidirectional transport of ATM cells), LLC/SNAP encapsulation is used. Under this scheme, the protocol is identified by prefixing the IP packet with a logical link control (LLC) header (defined by the IEEE 802.2), followed by a subnetwork attachment point (SNAP) header (defined by IEEE 802.1a). The SNAP consists of 2 subfields: A 3 octet Organisationally Unique Identifier (OUI). The bit pattern in the OUI determines the meaning in the following 2 octet Protocol Identifier (PID). Fig 2.3.1a illustrates an LLC/SNAP encapsulated packet.[5]

The maximum transfer unit (MTU) size is standardized at 9180 bytes, excluding the 8 byte LLC/SNAP header. However, the MTU size can be negotiated upto 64kbytes. LLC/SNAP encapsulation is the default scheme for IP over ATM.

In VC multiplexing, each protocol is carried over a separate VC with the protocol type specified at connection setup. As a result, minimal bandwidth & processing overhead are the benefits of this process.

IP datagrams contain the source & destination addresses. For an IP address, sometimes respective hardware address is unknown. Or in the case of IP over ATM, the destination's ATM address is unknown. In such cases, the Address Resolution Protocol (ARP) is used. The sender will broadcast an ARP request message to all stations. A station which recognizes its own IP address within this ARP request will send back an ARP reply containing its hardware address (and IP address). This enables the source to send the data to the designated destination. In brief, the ARP specified in RFC 826, binds a media access control (MAC) i.e. Hardware address to an IP address.

The classical view of an IP network is one in which clusters of IP nodes (end systems & intermediate systems ie. routers) with similar subnetwork addresses are connected externally to nodes out of their cluster by IP routers. RFC 1577 defines the "Classical IP over ATM" model [6]. The model groups IP nodes in an ATM network into logical IP subnets (LIS). The nodes in one LIS communicate to nodes in another thru an IP router. An ATM address resolution protocol server (ATMARP) is a component of each LIS. This performs the directory services function for the nodes in the LIS. Each node is configured with the ATM address of its ATMARP server. The address resolution process works in a client/server manner with the nodes as the clients.

When a node is activated, it establishes a connection with the ATMARP server (as its address is preconfigured) for its LIS. On detecting the connection from the new client, the ATMARP will then transmit an Inverse ARP [7] requesting the client's IP address since the client's ATM address is known thru the connection. After receiving the client's response, the ATMARP server stores the client's IP & ATM addresses in its ATMARP table.

Before a client can communicate with another, in the same LIS, the IP address of the latter must be resolved to its ATM address. To accomplish this, the client must transmit an ATMARP request to the ATMARP server. This will then respond with the target client's ATM address, if a match is found in the ATMARP table. Else an indication that the receiver is not registered will be the response. It is the responsibility of the client to regularly refresh its entry in the ATMARP using the procedure in the above paragraph, else its entry will be deleted.

Fig 2.3.2a shows Classical IP over ATM Architecture.

Fig 2.3.2a

The four basic types of equipment which make up an ATM network are:

1. Customer Equipment (CEQ)
2. ATM switches
3. ATM crossconnects
4. ATM multiplexors

Please refer to Appendix A for equipment prices [8].

 The following interfaces defined by ATM specifications form the basis for connections between the various components:

1. User Network Interface (UNI)

This is a technical specification which allows ATM customer equipment from various manufacturers to communicate over a network provided by yet another manufacturer. This is the interface used between ATM CEQ & either an ATM multiplexor, ATM crossconnect or an ATM switch.

2. Network Node Interface (NNI)

This is the interface employed between nodes within a network or between different subnetworks. A standardized NNI provides the scope to construct an ATM network from individual nodes supplied by different manufacturers.

3. Inter-Network Interface (INI)

The INI allows for communication between interconnected ATM networks. It's based upon the NNI but includes more features for ensuring security, control & proper administration.

[9]

The following list is a sample of some of the applications made possible or optimized by ATM

• Working at home
• Home shopping using voice, video & on-line databases
• Video on demand for popular movies
• Interactive multimedia games & applications
• Distance learning
• On-line video libraries
• Medical imaging
• Remote database access

One of the main applications expected to use ATM is Multimedia. The Internet was designed to handle the "best effort" delivery of IP traffic. Therefore delivery of voice, video & other real-time traffic is adversely affected as they demand guaranteed Quality of Service. With the arrival of low cost multimedia computers, & the availability of digital & audio applications, the need to introduce real-time traffic with guaranteed QoS is becoming important. New protocols are currently under development by the IETF to enable this service, referred to as Internet Integrated Services (IS). IS over ATM is also currently being addressed by the IETF. The benefits of ATM will be realised when its used to network multimedia workstations. Now, companies & other organisations want to gain a better visibility at the Internet & invest funds into upgrading the content of their websites. This is to deliver a full multimedia presentation to the observer. In the near future ATM is intended to be used as a backbone for other services like frame relay.

ATM is expected to improve LAN/Client Server architectures & LAN interconnection. It will provide resources to ease the burden on existing ones caused by the growth in the number of users, and allow for applications requiring greater bandwidth.

I have attempted to provide an investigation into "Internetworking over ATM" and closely related topics. The dissertation has been a good vehicle to learn how to conduct relevant research into an area which I was interested in. Also, organisational skills were of importance, especially in terms of time management as well as controlling the content of the report because of the word limit.

As network requirements increase coupled with a reduction in equipment prices, ATM is a technology which will prove to be popular & essential for the future.

http://www.cyberus.ca/~swanson/index.html

http://www.atmreport.com/

http://www.gmd.de/Topics/ATM/IETF/ipatm.html

http://www.uswest.com/pcat/small_business/ATM_Cell_Relay.html

http://www.atri.curtin.edu.au/crc/html/research.html

http://ipoint.vlsi.uiuc.edu/atm/atm.html

http://www.stel.com/stel/atmps/anma/anma.htm

http://www.eeng.dcu.ie/~murphyj/publ/ccsds-atm/ccsds-atm.html

http://sulu.lerc.nasa.gov/5610/atm.html

http://cell-relay.indiana.edu/cell-relay/

http://cell-relay.indiana.edu/cell-relay/FAQ/ATM-FAQ/d/d.htm

http://gump.bellcore.com:8000/~gja/interop95/interop95-p2.html

http://www.frforum.com/4000/fratm/fratm.toc.html

http://cell-relay.indiana.edu/cell-relay/publications/InternetResources/People.html

http://www.atmforum.com

http://www.infotech.tu-chemnitz.de/~paetz/atm/

BT Technology Journal Vol.16 No.4 October 1998

IEEE Journal on Selected Areas in Communications September 1998