ATM Tutorial: Transport Layer

I. Transport Layer in ATM

It is hard to say whether the ATM has a transport layer. On the one hand, the ATM layer has the functions of the network layer, and there is a layer of AAL on it). From the perspective of layering, AAL is the transport layer. Some experts agree. AAL5, one of the protocols used here, is similar to UDP, and UDP is undoubtedly a transport layer protocol. On the other hand, there is no AAL protocol that provides reliable end-to-end connections like TCP, although these protocols only require small changes ). In addition, in most applications, another transmission layer is used on AAL. I will not go into details. I will discuss the AAL layer and its protocol in this chapter, whether it is a real transport layer or not.

The AAL layer of the ATM network is essentially different from that of TCP because the designers are more interested in transmitting audio and video data streams. Therefore, fast transmission is more important than accurate transmission. The ATM layer outputs 53 bytes of cells consecutively. No error control, no traffic control, and other types of control in cells. Therefore, it cannot meet the requirements of most applications. To make up for this problem, in recommendation I .363, ITU defines an end-to-end Layer on the ATM layer. This layer is called the ATM adaptation layer (AAL). It has gone through a tortuous history of errors, revisions, and unfinished work.

The goal of AAL is to provide useful services to applications and separate them from those on the sender side) to split data into cells and separate them on the receiver side) and reorganize the cells as data. It organizes the service space according to the three axes:

1. Real-time and non-real-time services.
2. constant bit rate service and Variable Bit Rate Service.
3. connection-oriented and non-connection services.

In principle, eight different services can be defined using three coordinate axes and two values on each coordinate axis, for example. ITU considers that only four of them are useful and named as class A, Class B, Class C, and Class D respectively. Others are not supported. Since the ATM, this image is somewhat outdated, so it is proposed here as the background information to help readers understand why the AAL protocol is designed as it is currently. Currently, the main difference is between the transport class ABR, CBR, NRT-VBR, RT-VBR and UBR), rather than between the services supported by these AAL.

The basic services supported by AAL are outdated)

To deal with these four types of services, ITU defined four protocols and later found that the technical requirements for Class C and Class D are very similar, so that the combination of A_3 and AAL4 is A_3/4. The computer industry was not aware of it at the time, and later found that they were not satisfactory. Later, another protocol named AAL5 was defined to solve the problem.

Ii. Structure of the ATM Adaptation Layer

The above section of the ATM adaptation layer is called the aggregation sublayer (cenvergence sublayer ). It provides an interface to the application. It is composed of two sub-parts: one is the public part common to all applications relative to the given AAL Protocol), and the other is the sub-parts related to the application. Each part is related to the Protocol, but it can include packet frame splitting and error detection.

On the sender side, the convergence sub-layer is responsible for receiving the bit stream data from the application) or random length packets, and divides them into 44 ~ 48-byte unit for transmission. The exact size depends on the protocols used, because some protocols occupy part of the 48-byte ATM load as their own headers. At the receiving end, the child layer reassembles the cells into the original packets. Generally, if the message boundary exists, it must be retained.

The parts below AAL are called the segmentation and reassembly child layers. It adds the header and tail of the Data Unit that the convergence child layer handed over to it to form the cell payload. Then, these loads are handed over to the ATM layer for transmission. At the receiving end, the SAR Sub-layer reassembles cells into packets. The SAR Sub-layer basically only involves cells, while the convergence sub-layer deals with packets.

The SAR Sub-layer provides other functions for some but not all) Service classes. In particular, it can sometimes perform error detection and multiplexing. The SAR Sub-layer is applicable to all service classes, but its function strength depends on its specific protocols.

Iii. AAL1

AAL1 is A protocol used for Class A transmission. Class A transmission refers to real-time, constant bit rate, and connection-oriented transmission, such as non-compressed audio and video data. The input is a bit stream and there is no message division. For such transmission, there is no error detection protocol such as stop, because the delay introduced by the timeout and retransmission mechanisms is unacceptable. However, when a cell is lost, the application is notified to take measures if possible.

AAL1 uses a convergence sublayer and a SAR sublayer. The Convergence sub-layer detects lost and accidentally entered cells, flat the input data rate, and sends cells at a constant speed. Finally, the convergence sub-layer splits the input packets or bit streams into 46 bytes or 47 bytes, and then submits them to the SAR Sub-layer for processing. At the receiver end) It extracts the data units and restructured them into original input. The accesskey sub-layer of AAL1 does not have its own protocol header information.

On the contrary, the SAR Sub-layer of AAL1 has its own protocol. The cell format is as follows. Both formats start with a 1-byte header: The 3-byte cell SN is used to detect whether cells are lost or accidentally entered ); this field is followed by a three-digit serial number protection field SNP (checksum), which can correct a single error in the cell serial number field and detect two errors.

P cells are used when the message boundary must be retained. The Pointer field is used to indicate the offset of the starting position of the next packet.

Iv. AAL

AAL1 is designed for simple, connection-oriented, and real-time data streams. In addition to its detection mechanism for lost and accidentally imported cells, it has no error detection function. AAL1 is sufficient for pure uncompress audio or video data, or for any other data streams with occasional important bits.

For compressed audio or video data, the data transmission rate changes significantly over time. For example, many compression schemes periodically send complete video data when transmitting video data, then only send the difference between adjacent sequence frames, and finally send a complete frame. When the lens is still and nothing moves, the difference frame is very small. Second, the packet boundary must be retained to distinguish the starting position of the next full frame, even when cells or bad data are lost. For these reasons, a more comprehensive Protocol is required. AAL is designed for this purpose.

Like in AAL1, the convergence sub-layer of AAL 2 does not have its own protocol, while the SAR Sub-layer has its own protocol. The format of SAR cells is as follows:

AAL cell format

The SN (Sequence Number) field is used to record the Number of a cell to detect the loss or incorrect entry of the cell. The IT (Information Type) field indicates the start, center, or end of a message. The Length indicates the size of the payload in bytes, which may be less than 45 bytes ). Finally, the CRC field is the checksum of the entire cell and can detect errors.

The size of each field is not specified in the standard. It is said that at the last moment of the standardization process, the Committee members felt that there were many problems in Aal, so that they could not be put into use, but it was too late to organize the standardization process. Finally, the members removed all the field size settings so that the formal standards can be published on time, but no one can actually use it.

V. A_3/4

At the beginning, ITU developed different protocol service classes C and D for service classes C and D are sensitive to data loss or error, respectively, but it does not have a real-time connection-oriented and non-connection data transmission service class ). Later, ITU found that there was no need to specify two sets of protocols, so it merged them into one and formed a separate protocol, namely A_3/4.

A_3/4 can be operated in two modes: stream and message. The message boundary information is not retained in stream mode. The following describes the stream mode. In each mode, reliable transmission and unreliable transmission may occur, that is, the reliability is not guaranteed.

A_3/4 has a performance that is not available in other protocols-supports multiplexing. This feature of A_3/4 allows multiple sessions from a host, such as remote logon, to be transmitted along the same virtual circuit and separated from the target. All sessions using a virtual circuit get services of the same quality, because this is determined by the nature of the virtual circuit itself.

Unlike AAL1 and AAL 2, A_3/4 has the convergence sub-layer protocol and the SAR Sub-layer protocol. The maximum number of packets sent from the application to the convergence sub-layer is 65535 bytes. First, fill it with an integer multiple of 4. Add the header and tail information. The Convergence sub-layer reconstructed the heat preservation and added the header and tail information, then the packet is sent to the SAR Sub-layer, which divides the packet into a data segment of up to 44 bytes.

A_3/4 has two-layer protocol Overhead: each packet needs to be increased by 8 bytes, and each cell needs to be increased by 4 bytes. In short, it is a mechanism with high overhead, especially for short packets.

6. AAL5

The protocols from AAL1 to A_3/4 are mainly designed by the telecom industry and standardized by ITU. They do not have much consideration for the requirements of the computer industry. Due to the complexity and inefficiency caused by the two protocol layers, and the addition of the checksum field is very short and only 10 characters), some researchers have come up with the idea of creating a new adaptation layer protocol. The Protocol is called simple efficient adaptation layer (SEAL). After demonstration, the ATM Forum accepted SEAL and named it AAL5.

AAL5 provides several services to its applications. One option is reliable service, that is, the traffic control mechanism is used to ensure transmission to prevent overload), and the other option is unreliable service, that is, no data transmission guarantee measures are provided ), use the option to make the cells that are incorrectly verified or lost or transferred to the application but identified as bad cells ). AAL5 supports point-to-point transmission and multi-VOD Transmission. However, the multi-point transmission mode does not provide data transmission assurance measures.

Like A_3/4, AAL5 supports message mode and stream mode. In packet mode, the application can extend the length from 1 byte ~ A 65535-byte datagram is transmitted to the AAL layer. When the convergence child layer is reached, fill in the message to the payload field with the tail information, and select fill DATA 0 bytes ~ 47 bytes), so that the entire packet includes the filled data and tail information) is an integer multiple of 48 words. AAL5 does not converge the child layer header and only has an 8-byte tail.

The User-to-User field is not used in the AAL layer, but for your own purposes. A higher layer may be a specific service sub-division of the sub-layer. For example, sort or multiplexing. Length) indicates the actual payload, in bytes, excluding the number of bytes filled. The value 0 is used to terminate messages that have not been transmitted. The CRC field is based on the standard 32-bit checksum of the entire message, including the CRC Field set to 0 for the filled data and tail information ). An eight-digit field at the end is reserved for future use.

The packets are sent to the SAR Sub-layer and then sent out. In the SAR Sub-layer, no header or tail information is added. Instead, the packet is divided into 48 bytes and each unit is sent to the ATM layer for transmission. It also notifies the ATM layer to set the PTI field of the last cell to 1 so as to retain the packet demarcation. A problem occurs: This is an incorrect mix of protocol layers, because the AAL layer should not use the header information of the ATM layer .)

AAL5 is more efficient than A_3/4. Although A_3/4 only adds 4 bytes of header information for each packet, it also adds 4 bytes of header information for each cell, thus reducing the capacity of the payload to 44 bytes, for long packets, invalid data accounts for 8%. Each AAL5 packet has a slightly larger tail 8 bytes), but each cell has no additional overhead. The cell does not have a sequence number. It can be compensated by a long checksum to detect lost, mistakenly inserted, or wrong cells without the sequence number.

On the Internet, the common method for an ATM interface is to use the AAL5 payload field to transmit IP groups. Various issues related to this method are discussed in RFC 1483 and RFC 1577.

VII. Comparison of AAL protocols

The various AAL protocols seem to be unnecessary and have very poor consideration. It is also questionable to distinguish the convergence sub-layer from the SAR Sub-layer, especially because the SAR Sub-layer of AAL5 does not have any characteristics of itself. It is sufficient to use slightly enhanced ATM layer header information to provide image sorting, multiplexing, and data frame splitting functions.

The overall impression of AAL is that there are many variants, and there are many subtle differences between them, and they are not yet completed. The four original Services A, B, C, and D have actually been abolished. AAL1 may not exist; AAL 2 is incomplete; AAL and AAL4 are never at the beginning; AAL/4 is inefficient and there are too few checksum fields.

In the future, everything will depend on AAL5, but so far, there is still much room for improvement in AAL5. AAL5 packets should have a sequence number and a flag used to distinguish between data and control packets, thus becoming a reliable transmission protocol. You can use unused space at the end to implement the above functions.

VIII. SSCOP-connection-oriented protocol for specific services

Although there are so many different AAL protocols, there is no simple and reliable point-to-point transmission connection. Applications that require such services can use another protocol-connection-oriented SSCOP (service specific connection oriented protocol) for specific services ). However, SSCOP is only used for control and cannot be used for data transmission.

Each packet is assigned a 24-bit sequence number. Messages can be up to 64 KB and cannot be separated. They must be transmitted in order. Unlike some reliable transmission protocols, when a packet is lost, it is always selectively re-transmitted instead of returning all packets after the serial number n.

SSCOP is essentially a Dynamic Sliding Window Protocol. For each connection, the receiver retains the window for receiving the serial number of the message and specifies whether the bitmap of the message already exists ). This window can be changed during protocol operations.

What is unusual about SSCOP is the way to handle validation: it does not have a tape mechanism. Instead, the sender periodically queries the receiver and requests it to send a bitmap indicating the window status. Accordingly, the sender discards the packets received by the other party and updates its window.