Fundamentals of AAL2
The AAL2 protocol has two layers:
Service specific convergence sublayer (SSCS)
Common part sublayer (CPS)
The SSCS encodes different information streams for the transport by AAL2 over a single ATM connection. The information streams might be active voice encodings, silence insertion descriptors, dialed digits, or fax. SSCS can provide error control on critical information (CAS signaling and dialed digits) by using a 10-bit CRC. This is called an AAL2 Type 3 cell. The SSCS segments the information that is being passed from a higher layer application, such as samples of voice from a digital phone into a number of units of data, and submits these units of data to the CPS for transmission. The length of the segmented data can be between one and the maximum length supported by the CPS connection, which is either 45 or 64 bytes. At the SSCS receiver, the units of data are reassembled back into the information before being passed to the higher layer application.
The second layer, the CPS, is specifically responsible for transporting end-to-end connections across the network. The format of AAL2 protocol structure is shown in Figure 4-2.
Figure 4-2 AAL2 Protocol Structure
Figure 4-2 shows that AAL2 uses an additional byte of overhead for each ATM cell and an additional three bytes of overhead for each voice packet (e.g., compressed 8 kbps voice). The benefits of the AAL2 scheme are that there is no padding overhead except when there is insufficient data to complete a packet in a prespecified time interval, and the voice channels can be multiplexed over a single ATM virtual circuit.
The content of the Start field and the CPS packet are shown in Figure 4-3.
Figure 4-3 Start Field and CPS Packet Formats
The CPS layer enables the multiplexing of variable length voice packets of end users onto a single ATM virtual channel that is an AAL2 channel. This is accomplished through the different information fields shown in Figure 4-2. Although AAL2 with its three-byte packet header introduces some inefficiency for small packets, the improvement that is reached by having no padding more than offsets this minor inefficiency. Each of the CPS fields and the Start field are described here:
Start fieldEnables efficient packing of the voice packets over a single ATM virtual circuit. The Offset field is a six-bit pointer within the Start field that points to the position of the first CPS packet that follows the OSF. A sequence number protects the order of the Offset field. If a Start field parity error exists, all the CPS packets that are associated with the Start field are discarded.
CID (Channel ID)Identifies the end user, which is referred to as the SSCS entity in the International Telecommunication Union (ITU) AAL2 specifications. The CID allocates the value 1 to exchange layer management peer-to-peer procedures, such as set-up negotiations. CID enables the multiplexing of up to 248 user channels, whereas some CID values are reserved for other uses, such as peer-to-peer layer management. For example, if 8 E1s terminated on an MGX, 240 CID values would be used.
LI (Length Indicator)Identifies the length of the CPS packet. The default payload length is 45 bytes, and an optional maximum length of 64 bytes can be selected. The maximum length is channel specific.
UUI (User-to-User Indication)Provides two functions: It conveys specific information transparently between two end points (e.g., CPS or SSCS entity) and distinguishes between the different users, such as SSCS entities and layer management users.
HEC (Header Error Control)Discards the rest of the CPS packets until the next Start field. As a result, not all voice users residing on the single ATM virtual channel are affected by other end-user errors, which results in a higher end-to-end efficiency.
The CID is an important concept in AAL2. CIDs provide a binding between an endpoint and an AAL2 connection. This is the mechanism that binds the TDM traffic to the ATM traffic. For example, if a service provider needs to provision 100 DS0s between two sites for one of its enterprise customers, 100 CIDs are created across the ATM network. Furthermore, a unique coder-decoder (codec) type is assigned to each individual DS0 because the codec type is assigned to each CID through an AAL2. For example, individual customers in a multitenant building can each support multiple compression schemes over a single T1 access link. Each CID is configured and includes the following parameters: codec type, profile type, voice activity detection (VAD), DTFM Tones, and packet period for G.729. For example, to transmit DTMF tones transparently across the ATM PVC, DTMF must be enabled in the CID.
An AAL2 profile is a mechanism that the MGX 8850 uses to assign the compression and encoding scheme of the AAL2 trunking service. A profile is defined by a profile type, which is either an ITU standard or a custom type and a number. These profiles need to match on both ends of the network for the two end devices, such as PBXs, to interoperate. A profile is configured for each CID. For example, if the profile type is ITU and the profile number is 1, you must use G.711. In other words, the profile type and the profile number identify the compression type.
CID enables the use of subcell multiplexing, which provides many of the benefits of AAL2. If you use G.711, subcell multiplexing does not provide any value because G.711 already uses an 80-byte packet. The real advantage of subcell multiplexing is the G.729 encoding scheme. If you use G.729 with a packetization period of 30 milliseconds, three 10-byte packets of payload from one DS0 are packed into one ATM cell. Therefore, the efficiency of packing the voice sample into the ATM cell is increased threefold, and instead of 34 bytes of padding, only 14 bytes exist in the ATM cell. Assuming that VAD provides an additional 50 percent of bandwidth savings, G.729 subcell multiplexing uses approximately 6 kbps of bandwidth per DS0 channel of voice traffic. This is a significant amount of bandwidth savings.