Introduction to Cell Switching and ATM
In the ATM model, cell switching combines a connection-oriented logical path with the transfer of short data units. It is a variant of packet switching where the data packets, called cells, have a fixed, small size. It was designed to handle different types of traffic, from delay-sensitive voice and video to bursty computer data.
The best-known technology that implements cell switching is ATM (Asynchronous Transfer Mode).
Why Use Small, Fixed-Size Cells?
The choice of small, fixed-length cells was an engineering response to a central challenge of integrated networks: how to provide quality for real-time services while carrying data traffic.
- Predictable Delay (Latency): In a traditional packet network with variable packet sizes, a small urgent voice packet might wait behind a large file-transfer packet. The serialization time of the larger unit increases delay and its variation for voice traffic. A fixed-size cell limits this component of delay, although queuing and congestion can still add delay.
- Hardware Switching: The fixed and simple structure of cells simplifies fast hardware processing. Cell boundaries follow the fixed length and physical-layer procedures, so the device does not need to handle a variable length for every unit. This enabled links with rates of hundreds of Mb/s and above.
- Single Technology for All Services: ATM was designed as the transfer mode for , a single network for all services. Cells provided a compromise: their serialization time was limited enough for voice, and adaptation layers allowed them to carry data.
Anatomy of an ATM Cell
An ATM cell has a total length of 53 bytes. This size reflects a compromise between limiting delay and keeping header overhead low. The cell is divided into a 5-byte control header and a 48-byte user-data payload.
Header Fields Explained
- GFC (Generic Flow Control): A 4-bit field present only at the User-Network Interface (UNI). It was intended for traffic-flow management, but had little practical role in many deployments.
- VPI/VCI (Virtual Path/Channel Identifier): These values are local to an interface, and a switch or cross-connect can translate them at a link boundary. At the UNI, VPI has 8 bits, at the NNI it has 12 bits, and VCI has 16 bits. ATM uses a two-level hierarchy: multiple Virtual Channels (VCs, identified by VCI) can be bundled into a Virtual Path (VP, identified by VPI). A VP switch forwards the bundle using VPI, while a VC switch also uses VCI.
- PT (Payload Type): Indicates whether the payload carries user information or management information. For user cells, PTI also carries congestion and user-to-user indications; other codes identify OAM F5 and resource-management cells.
- CLP (Cell Loss Priority): A single bit that sets the cell's loss priority. If , the cell is considered low-priority and may be discarded by the network during congestion. If , the network gives it higher priority, but delivery is still not guaranteed.
- HEC (Header Error Control): The 8-bit HEC is generated from the first four header octets and protects the 40-bit header. The code supports single-bit correction and multiple-bit error detection. In detection mode, cells with detected header errors are discarded, limiting the risk of misrouting after VPI/VCI corruption. Error control for the 48-byte payload is handled by higher layers.
ATM's Legacy and Influence
ATM found widespread use in networks and switch designs, although it did not become the universal B-ISDN technology originally envisioned.
- Use in LANs: In local area networks, ATM was displaced by Ethernet, which offered simpler and less expensive infrastructure and evolved with enterprise requirements.
- Use in Core Networks: For many years, ATM was widely used in telecom-operator backbone networks. It aggregated IP, voice, and Frame Relay traffic on a single high-capacity infrastructure.
- Influence on MPLS: ATM's use of short, fixed-length labels (VPI/VCI) for fast hardware switching decisions influenced the development of , which is used in carrier and ISP networks.