ATM Architecture

The Core Components of an ATM Switch

An ATM switch is a device used in an ATM network. Its purpose is to forward fixed-size data units, known as ATM cells, from an input port to the correct output port. A switch typically consists of N input ports and N output ports, all interconnected by a central switching element.

At the UNI, a switch connects to an end system; at the NNI, it connects to another network node. The UNI header contains a 4-bit GFC field for generic flow control. The NNI format has no GFC field, and those four bits extend the VPI field. At UNI, VPI is 8 bits and VCI is 16 bits; at NNI, VPI is 12 bits and VCI remains 16 bits.

The architecture is modular and includes four main components:

• Input modules (line cards): Receive the incoming signal and prepare cells for switching.
• Output modules (line cards): Prepare cells from the fabric for onward transmission.
• Switching fabric: The element that physically transfers cells from inputs to outputs.
• Control and management unit: Handles connection management and switch operation.

ATM layers and planes

The B-ISDN (Broadband Integrated Services Digital Network) reference model separates the user, control, and management planes. The user plane transfers user information, the control plane handles signaling and connection setup and release, and the management plane covers layer management and management functions for the system as a whole.

The functional model distinguishes the Physical Layer, ATM Layer, and ATM Adaptation Layer (AAL). The Physical Layer includes the Physical Medium (PM) and Transmission Convergence (TC) sublayers, while the ATM Layer separates virtual path (VP) and virtual channel (VC) functions. AAL adapts higher-layer service data for transport in ATM cells. SAAL (Signaling ATM Adaptation Layer) includes AAL functions used to transport signaling.

An example cell path runs from signal conversion at the input, through VPI/VCI lookup and the switching fabric, to the output buffer, where the header is translated and HEC recalculated.

The Input Module

The input module, typically part of a line card, performs a set of tasks to process the incoming signal and prepare ATM cells for the switching fabric.

1. Signal reception and conversion: Receives the physical optical or electrical signal from the line.
2. Physical layer processing: In systems like SDH/SONET, this involves bit and frame synchronization, processing the transport frame header, and extracting the ATM cell stream from the frame's payload.
3. Cell delineation: Analyzes the bitstream to identify the boundaries of individual ATM cells that are 53 bytes long.
4. Header validation (HEC): Checks each cell header using the Header Error Control (HEC) field. HEC can correct a single-bit error or detect multiple-bit errors.
5. Cell discarding: Does not pass idle cells to the ATM layer and discards cells with uncorrectable header errors.
6. Traffic policing (UPC/NPC): Usage Parameter Control (UPC) monitors traffic at the user-network interface (UNI), while Network Parameter Control (NPC) monitors traffic at the network-node interface (NNI). Both check traffic against the negotiated parameters and may discard or mark cells that violate the traffic contract.
7. Forwarding table lookup: The VPI/VCI values from the cell header are used as a key in a forwarding table to determine which output port the cell should be sent to.
8. Internal header prepending: An implementation may add an internal header or tag. The switching fabric uses it to forward the cell to the appropriate output module, which then removes the tag.

The Switching Fabric

The switching fabric is the switch element responsible for physically transferring cells from the input modules to the designated output modules.

• Synchronous operation: The fabric typically operates synchronously. Time is divided into equal intervals called "cell slots," with each slot corresponding to the time it takes to transmit one ATM cell.
• Capacity requirement: The fabric's capacity must match port loads and the buffering strategy. For N ports with the same line rate, their total maximum rate can reach N times the rate of one line. Capacity and headroom depend on the architecture, buffering strategy, and traffic model. When traffic from multiple inputs converges on one output, the requirements also depend on traffic concentration.
• Performance measures: Its performance is measured by parameters like normalized throughput and utilization.

The Output Module and Control Unit

The output modules prepare cells for transmission to the next node, while the control unit manages the process.

Output module functions

• Cell buffering: Incoming cells from the fabric are temporarily stored in a buffer to await transmission.
• Header translation: The module replaces the previous VPI/VCI values in the cell header with new ones appropriate for the next link and recalculates the Header Error Control (HEC) field.
• Physical transmission: The processed cell is handed over to the physical layer unit, which converts it into an optical or electrical signal and inserts it into the outgoing transport frame (e.g., SDH).

Control and management unit

This unit performs the tasks needed to establish and manage connections:

• Exchange signaling information: The switch must communicate with other network nodes to set up connections.
• Resource allocation: It must make decisions about allocating resources (such as bandwidth and buffer space) for new connection requests.
• Path selection: It must choose the appropriate route through the network for a new connection.
• Admission control (CAC): Decides whether to accept or deny new connection requests to prevent network overload.