ATM Adaptation Layer type 2

Description

ATM Adaptation Layer type 2 (AAL2) is a critical protocol defined by the ITU-T and adopted by 3GPP for the transport of real-time, variable-bit-rate traffic in the early releases of the UMTS network architecture. It operates as a sublayer within the ATM protocol stack, sitting between the service-specific higher layers (e.g., voice codecs, signaling protocols) and the core ATM layer. Its primary function is to segment and reassemble the data streams from multiple users or applications into the fixed-size cells (53-byte) used by ATM, while preserving timing relationships and minimizing packetization delay, which is essential for conversational services.

Architecturally, AAL2 is divided into two sublayers: the Common Part Sublayer (CPS) and the Service Specific Convergence Sublayer (SSCS). The CPS is the core, responsible for multiplexing multiple AAL2 connections, called Channel Identifiers (CIDs), into a single ATM Virtual Channel Connection (VCC). It does this by packing variable-length CPS-Packets from different sources into the 48-byte payload of ATM cells. Each CPS-Packet consists of a 3-byte header (containing the CID, Length Indicator, and User-to-User Indication) and a payload of up to 45 or 64 bytes (depending on the mode). This packing mechanism, known as mini-cell multiplexing, allows for very low delay as cells can be sent before being completely filled, unlike AAL1 or AAL5, making it highly efficient for sporadic, low-bit-rate traffic.

Within the 3GPP context, AAL2 was specified as the primary transport for user plane traffic (e.g., AMR voice, circuit-switched data) and control plane signaling across critical interfaces in the UTRAN. Specifically, it was used on the lub interface between the Node B and the Radio Network Controller (RNC), the lur interface between RNCs, and the lu-CS interface between the RNC and the Core Network's Circuit-Switched domain (MSC). The SSCS part of AAL2 was further specialized for these roles; for example, the Service Specific Connection Oriented Protocol (SSCOP) and Service Specific Coordination Function (SSCF) were used for signaling transport, while a null SSCS or a segmentation and reassembly SSCS was often used for user data. This layered approach provided a reliable, low-latency pipe that was fundamental to the Quality of Service (QoS) guarantees of early 3G services.

AAL2 works by establishing AAL2 connections end-to-end between network elements. When a user plane data packet (like a voice frame) arrives from the higher layer, the AAL2 layer at the source (e.g., Node B) encapsulates it into a CPS-Packet with a specific CID corresponding to that user's bearer. Multiple such CPS-Packets from different CIDs are then sequentially packed into the payload of an ATM cell. The cell is transmitted once a timer expires or the payload is nearly full, ensuring a compromise between delay and efficiency. At the receiving end (e.g., RNC), the CPS extracts the CPS-Packets based on the CID and length indicator, reassembles the original data units, and delivers them to the correct higher-layer entity. This process, managed by the AAL2 signaling protocol (Q.2630.1/Q.2150.1 in 3GPP), allowed for dynamic establishment and release of these low-bandwidth channels, aligning perfectly with the call setup and release procedures in UMTS.

Purpose & Motivation

AAL2 was created to address the specific challenge of efficiently transporting bursty, low-bit-rate, and delay-sensitive traffic over ATM networks, which were the dominant high-speed backbone technology in the late 1990s and early 2000s. Prior approaches like AAL1 were designed for constant-bit-rate traffic (e.g., E1/T1 circuit emulation) and were inefficient for voice, which has silent periods, while AAL5 was optimized for large, bursty data packets but introduced significant packetization delay unsuitable for real-time communication. The telecommunications industry, moving towards integrated networks, needed a method to carry compressed voice (e.g., using AMR codecs) and other real-time services without wasting the fixed bandwidth of ATM cells.

In the historical context of 3GPP Release 99, which standardized the first UMTS networks, ATM was selected as the transport technology for the new UTRAN due to its proven reliability, traffic management, and QoS capabilities. However, to make this viable for the myriad of low-rate radio access bearers, an adaptation layer was required that could statistically multiplex hundreds of voice calls onto a single ATM VCC while maintaining strict delay bounds. AAL2 solved this by allowing partial filling of cells and interleaving packets from multiple sources, dramatically improving bandwidth utilization compared to dedicating a full VCC per call. This efficiency was economically critical for mobile operators deploying 3G infrastructure.

Thus, the primary motivation for adopting AAL2 in 3GPP was to enable cost-effective, high-quality transport for circuit-switched services over a packet-oriented ATM core, bridging the gap between traditional telephony requirements and modern packet-switched backbone networks. It provided the necessary traffic engineering to support the QoS classes defined for UMTS, particularly the conversational and streaming classes, until the industry's eventual migration towards all-IP transport with technologies like IP over Ethernet in later releases.

Key Features

• Statistical multiplexing of multiple low-bit-rate connections (CIDs) into a single ATM VCC
• Low packetization delay via partial cell fill and timer-based transmission
• Support for variable-size payloads (up to 45/64 bytes) per CPS-Packet
• 3-byte header per CPS-Packet containing Channel Identifier (CID) and length
• End-to-end signaling (Q.2630.1) for dynamic connection establishment and management
• Segmentation and Reassembly capability for service data units larger than CPS-Packet payload

Evolution Across Releases

Defining Specifications