ATM is a high-speed networking standard designed to support both voice and data communications. ATM is normally utilized by Internet service providers on their private long-distance networks. ATM operates at the data link layer (Layer 2 in the OSI model) over either fiber or twisted-pair cable.
ATM differs from more common data link technologies like Ethernet in several ways. For example, ATM utilizes no routing. Hardware devices known as ATM switches establish point-to-point connections between endpoints and data flows directly from source to destination. Additionally, instead of using variable-length packets as Ethernet does, ATM utilizes fixed-sized cells. ATM cells are 53 bytes in length, that includes 48 bytes of data and five (5) bytes of header information.
The performance of ATM is often expressed in the form of OC (Optical Carrier) levels, written as "OC-xxx." Performance levels as high as 10 Gbps (OC-192) are technically feasible with ATM. More common performance levels for ATM are 155 Mbps (OC-3) and 622 Mbps (OC-12).
ATM technology is designed to improve utilization and quality of service (QoS) on high-traffic networks. Without routing and with fixed-size cells, networks can much more easily manage bandwidth under ATM than under Ethernet, for example. The high cost of ATM relative to Ethernet is one factor that has limited its adoption to "backbone" and other high-performance, specialized networks.
dvantages of the Technology
The use of cell-switching technology in a LAN environment provides some distinct advantages over the shared-medium technology employed by Ethernet, token-ring, and FDDI networks. Two of those advantages are obtaining full bandwidth access to ATM switches for individual workstations and enabling attaching devices to operate at different operating rates. Those advantages are illustrated in Figure 14.2, which shows an ATM switch that could be used to support three distinct operating rates. Workstations could be connected to the switch at 25Mbps, and a local server could be connected at 155Mbps to other switches either to form a larger local LAN or to connect to a communications carrier's network via a different operating rate.
The selection of a 53-byte cell length results in a minimum of latency in comparison to the packet length of traditional LANs, such as Ethernet, which can have a maximum 1526-byte frame length. Because the ATM cell is always 53 bytes in length, cells transporting voice, data, and video can be intermixed without the latency of one cell adversely affecting other cells. Because the length of each cell is fixed and the position of information in each header is known, ATM switching can be accomplished via the use of hardware. In comparison, on traditional LANs, bridging and routing functions are normally performed by software or firmware, which executes more slowly than hardware-based switching.
The actual routing of ATM cells depends on whether a connection was pre-established or set up as needed on a demand basis. The pre-established type of connection is referred to as a Permanent Virtual Connection (PVC), and the other type is referred to as a Switched Virtual Connection (SVC). Examine the 5-byte ATM cell header shown in Figure 14.1 and note the VCI and VPI fields. The VPI is 8 bits in length, whereas the VCI is 16 bits in length, enabling 256 virtual paths of which each path is capable of accommodating up to 65,536 (216) virtual connections.
By using VPs and VCs, ATM employs a two-level connection identifier that is used in its routing hierarchy. A VCI value is unique only in a particular VPI value, whereas VPI values are unique only in particular physical links. The VPI/VCI value assignment has only local significance, and those values are translated at every switch a cell traverses between endpoints in an ATM network. The actual establishment of a virtual path is based on ATM's network management and signaling operations. During the establishment of a virtual path routing table, entries in each switch located between endpoints map an incoming physical port and a Virtual Path Identifier pair to an outgoing pair. This initial mapping process is known as network provisioning, and the change of routing table entries is referred to as network reprovisioning.
Figure 14.3 illustrates an example of a few possible table entries for a switch, where a virtual path was established such that VPI=6 on port 1 and VPI=10 on port 8, representing two physical links in the established connection.
Figure 14.3: Switch operations based on routing table entries.
Next, examine the entries in the routing table shown in Figure 14.3, and note that the table does not include values for VCIs. This is by design because a VP in an ATM network can support up to 65,536 VC connections. Thus, only one table entry is required to switch up to 65,536 individual connections if those connections all follow the same set of physical links in the same sequence. This method of switching, which is based on the VPI and port number, simplifies the construction and use of routing tables and facilitates the establishment of a connection through a series of switches. Although VCIs are not used in routing tables, they are translated at each switch. To help you understand the rationale for this technique, you must focus on their use. As previously noted, a VCI is unique within a VP and is used at an endpoint to denote a different connection within a virtual path. Thus, the VPI/VCI pair used between an endpoint and a switch has a local meaning and is translated at every switch; however, the VCI is not used for routing between switches.
The establishment of a connection between two end stations is known as a Virtual Channel Connection (VCC). To illustrate the routing of cells in an ATM network based on a VCC, consider Figure 14.4, which represents a small two-switch–based ATM network. The VCC represents a series of virtual channel links between two ATM endpoints. In Figure 14.4, one VCC could be represented by VCI=1, VCI=3, and VCI=5, which collectively form a connection between workstations at the two endpoints shown in the network. A second VCC could be represented by VCI=2, VCI=4, and VCI=6. The second VCC could represent the transportation of a second application between the same pair of endpoints or a new application between different endpoints served by the same pair of ATM switches.
Figure 14.4: Connections in an ATM network.
As indicated by the previous examples, each VC link consists of one or more physical links between the location where a VCI is assigned and the location where it is either translated or removed. The assignment of VCs is the responsibility of switches during the call setup process.
Top of page
The ATM Protocol Reference Model
Three layers in the ATM architecture form the basis for the ATM Protocol Reference model, illustrated in Figure 14.5. Those layers are the Physical layer, the ATM layer, and the ATM Adaptation layer.
Figure 14.5: The ATM protocol suite.
The Physical Layer
As indicated in Figure 14.5, the lowest layer in the ATM protocol is the Physical layer. This layer describes the physical transmission of information through an ATM network. It is not actually defined with respect to this new technology. The absence of a Physical layer definition results from the design goal of ATM to operate on various physical interfaces or media types. Thus, instead of defining a specific Physical layer, ATM depends on the Physical layers defined in other networking protocols. Types of physical media specified for ATM include shielded and unshielded twisted-pair, coaxial cable, and fiber-optic cable, which provide cell transport capabilities ranging from a T1 rate of 1.544Mbps to a SONET range of 622Mbps.
The ATM Layer
The ATM layer represents the physical interface between the ATM Adaptation layer (AAL) and the Physical layer. Thus, the ATM layer is responsible for relaying cells from the AAL to the Physical layer for transmission, and in the opposite direction from the Physical layer to the AAL for use in an endpoint. When transporting cells to the Physical layer, the ATM layer is responsible for generating the five-byte cell header for each cell. When receiving cells from the Physical layer, the ATM layer performs a reverse operation, extracting the five-byte header from each cell.
The actual manner by which the ATM layer performs its relaying function depends on the location of the layer at a switch or at an endpoint. If the ATM layer is located in an endpoint, it receives a stream of cells from the Physical layer and transmits either cells with new data or empty cells if there is no data to send to the AAL. When located in a switch, the ATM layer is responsible for determining where incoming cells are routed and for multiplexing cells by placing cells from individual connections into a single-cell stream.
The functionality of an LEC is typically located in an ATM adapter card installed in a legacy switch. That card is configured with two addresses: an IEEE 48-bit MAC address and a 20-byte ATM address. The LEC is responsible for address resolution, data forwarding, and registration of MAC addresses with the LANE server (LES). It also communicates with other LECS via ATM virtual channel connections established across the ATM network.
The LANE Configuration Server maintains a database of emulated LANs (ELANs) and the ATM address of LAN Emulation Servers (LESs) that control the ELANs. When a LANE client needs an ATM address, it first searches its connections, called Virtual Channel Connections (VCCs), it previously opened. The LEC maintains a translation table of destination MAC addresses mapped to VCCs. If the destination address is in the table, the LEC can use the existing VCC to send the message. If not, the LEC must perform an address resolution procedure using the LAN Emulation Address Resolution Protocol (LE-ARP). To do so, it queries the LECS, which returns the ATM address that serves the appropriate emulated LAN. The LEC then uses that address to query the LES. The LECS database is defined and maintained by the network manager or LAN administrator and represents the only manual process in the entire emulation process.
The LES represents a central control point for a predefined group of LECs. The LES maintains a point-to-multipoint Virtual Control Channel to all the LECs it controls. When the LEC queries the LES, the LES verifies that the LEC can joint the ELAN. Assuming it can, it examines the request of the LEC to resolve a MAC to ATM address by searching its tables for the appropriate ATM address that provides a path to the desired MAC address. Those tables are formed by LECs registering their ATM-to-MAC address translations with the LES. If the address is in the LES' cache memory, the LES returns the ATM address to the LEC that uses that address to establish an ATM connection. If the LES does not have that address in cache memory, it uses the services of the BUS.
The Broadcast and Unknown Server (BUS) functions as a central point for transmitting broadcasts and multicast messages. It is required because ATM is a point-to-point connection-oriented technology that lacks a broadcast or one-to-many transmission capability. If the LES does not have the address required by the LEC, it uses the services of the BUS. That is, the BUS transmits an address resolution request to all stations that make up the ELAN, and the station that recognizes its own MAC address returns its ATM address. The LES updates its cache memory and returns the ATM address to the LEC. The LEC can then establish a connection across the ATM _network.
Although communications carriers have expended a significant amount of effort to develop an ATM infrastructure for transporting information between carrier offices, the expansion of this evolving technology to customer premises—as well as its common use on LANs—will probably take several years. This is because, as with any new technology, the cost of ATM equipment is relatively high in comparison to the cost of older technology. Over the next few years, you can expect several important standards to be promulgated, and you can also expect to see the cost of ATM equipment become more reasonable as development costs are amortized over a larger base of products. As this occurs, the use of ATM will expand considerably.