webhome - jay stuffs

Jay Suttiruttana
September 8, 1999

ATM developed in 1983 by AT&T Bell Labs as a data link layer protocol covering portion of the layer similar to LANs MAC layer minus the physical layer. It covered by SONET/DS3 for wide area communication or by other encoding/transmission methods (UTP, TAXI, or FCS) for local-area communication. ATM supported data rates up to 622 Mbs. It took nearly 10 years before an official forum was created to try to mature the product into production technology topology. By late 1980's some of the major characteristics of ATM were:

(LAN, ATM, and LAN EMULATION TECHNOLOGIES, Daniel Minoli and Anthony Alles: pp129-30):

• The decision to use connection-oriented methods rather than connectionless method. This implies that machinery of virtual circuits (VC) must be employed. Permanent virtual circuits (PVC) and switched virtual circuits (SVC) are also referred as permanent/switched virtual connections.
• The concepts of using asynchronous bandwidths allocation and transport, specifically ATM, rather than time-division bandwidth allocation and transport, as the underlying infrastructure. The concepts of Cells, their length and their characteristics, are simply a corollary of the decisions.
• The concepts of using an out-of-band signaling mechanism to set up VCs.
• The goal of making ATM a multi-service platform that can support a variety of fast packet and other services (e.g. frame relay service, SMD, private line emulation and carriage).

Early 1990's ATM focused its attentions on covering more media than just the data. Its development grew toward video, multimedia, and voice media. These media demanded network capable of handling large amount of data, and ATM suited for the environment. Developers wanted ATM to be more cost effective as a LAN technology with regard to access to desktop PC through twisted pair technology and making local ATM behave as LAN.

The first official release for ATM LAN Emulation (LANE) was in June 1995. November 1996, LANE 2.0 was released but found to lack many needed features such as scalability and multi-protocol support. LANE 1.0 compliant hardware was incapable of supporting LANE 2.0. A lot of expensive hardware from LANE 1.0 had to be replaced to keep ATM current LANE 2.0.

How Does ATM Work (Switching)?

ATM is an end station centric topology, which means a connection of VC must be established between the source and destination stations before data transfer. Unlike the Ethernet, which transmits the data onto the wire and rely on the networking hardware to route the information to its destination networks.

ATM uses permanent virtual connections (PVC) and switched virtual connections (SVC) between communicating station. (Virtual connection is a logical communication channel between end stations. Logical because this circuit is created along shared media that many also contain virtual connections providing a circuit between other end stations.) While the circuit must share available bandwidth along the media, communications are kept separate from each other through the use of connection identifiers.

With ATM connectivity, devices on the network called ATM switches maintain tables with the identifications of all end station. When a station needs to transmit data, it issues a maintenance packet call virtual path identifier (VPI) that propagates out into the network, setting up a virtual connection between two systems. VPI create a circuit-switched connection between two systems and ensure that each portion of the path along the way has enough available bandwidth to carry the signal. Once the circuit is complete, data transmission can begin. Figure 1 shows the virtual circuit and virtual path switching use by ATM. 

VPI can reserve the required bandwidth and only pick routes that can support the required transmission rate. This is a form of automatic load balancing of the network, which ensure the quality of service (similar to public telephone network (PTN) connectivity).

How Does ATM Function in Large Network?

Now that we have some understand on ATM internal connectivity. We can now see how ATM benefits large network in a broader point of view.

ATM/cell relay service is a technology that combines the concepts of frame relay and SDMS packet technology. In the late 1980's, a successful switched communications carrier services called frame relay and SMDS emerge. Figure 2 (a) shows an example of enterprise network composed of dedicated lines and the difficulty of adding a new site to the network and (b) enterprise using public switched services. These technologies employ the powerful switches developed to support switching functions. Their concepts of switching network proof its viable mechanism to support enterprise networking at WAN level. Unfortunately there are limitations to their services. SMDS support high speed DS3, but do not support voice and video in which the ATM heavily focused on. Frame relay only support access speed up to DS1. ATM takes the SMDS access speed and the frame relay media support capability and combined these into its own technology.

Router networks

Adding new site to router networks.

Figure 2a. Adding a new site to enterprise network composed of dedicated lines.

ATM network

Adding new site to ATM network

Figure 2b. Adding a new site to enterprise network using public switching services.

As you can see from figure 2b, ATM is ideal candidate for large networks. It can carry voice and data over network wire or fiber. ATM transmits all packets as 53 byte cells that have a variety of identifier on them to determine quality of service. It has the ability to leverage the largest existing network in the world (PTN) by integrating, without translation or modification. The connection can be made to LAN to MAN to WAN using strictly ATM. The PTN becomes extension of the local network, as no translation is required.

The main design goal of ATM is to support a wide range of data, video and voice applications in the same public network at a high speed. The key element of service integration is the provision of range of services using limited number of connection types and multipurpose user-network interfaces (UNI). ATM supports both on demand SVC connection and non-switched PVC connection. PVC may be used between fixed users and locations, while SVC affords maximum flexibility in establishing dynamic connections.

Unfortunately ATM wants to handle all of the end to end connections (same type of functionality provided by existing upper-layer protocols line IPX and IP), thus making configuration incorporated ATM functionality into an existing network infrastructure difficult.

Node maps, allows VPI to create path from one end station to another, maintains by ATM currently has no auto-discovery method for creating these tables in PVC circuits. The table must be maintains manually creating a nightmare situation when maintaining a network with many nodes. There is no guarantee that the circuit will be able to allocate the full bandwidth required by the end station if a WAN connection is required. This is similar to making a telephone call at peak hour and receiving a busy signal. If the connection is not permanent, there is no guarantee it will be there 100% of the time. In a network environment, that required connection to be make 100% of the time (e.g., banking, credit card transaction) this type of unpredictability is unacceptable.

Many large developer firms including Fore System and IBM have both invested heavily in the ATM-to-the-desktop technology (using ATM to link servers and workstations) and improving ATM's capabilities. Still ATM equipment is extremely expensive and esoteric. ATM itself is more complex than either Ethernet or Token Ring.