Asynchronous Transfer Mode

Asynchronous Transfer Mode (ATM) is a communications protocol which is suitable for the transmission of data, voice and video. The Layer 2 packets are called cells or slots have a fixed length ( 53 bytes, 48 bytes data, 5 bytes of cell header ) and transmitted via asynchronous time-division multiplexing.

  • 5.1 use in broadcasting and broadcasters
  • 5.2 Current Situation

Overview and history

ATM was developed as a network standard, the synchronous ( Plesiochronous Digital Hierarchy (PDH ), Synchronous Digital Hierarchy (SDH ) ) transport techniques used and adds them more useful features. Not only circuit-switched data transmission is supported by ATM, but also packet-based such as IP, Frame Relay, etc. In contrast to the simple and robust Ethernet technology, which can result in overload situations to unpredictable results, provides ATM guarantees regarding effective bit rate, delay and jitter, which commonly referred to ( among other properties ) the Quality of Service ( QoS).

The problem to combine a variety of data streams of different types, that is, for example, with both synchronous and packet-based networks to work together was achieved by both bit stream types (synchronous or packet-based) at the intersection points to a new bitstream with ATM cells are implemented. The cells are typically sent in the payload of PDH or SDH formatted data stream. Asynchronous in ATM means that the sender and receiver can divergent clock rates work within wide limits: The receiver checks by Header Error Check ( HEC) for each cell new, whether a sufficient synchronization exists, and if necessary performs a resynchronization by.

In the original design of ATM was the key technique for the ' Broadband Integrated Services Digital Network ' (broadband ISDN) which provides for the existing analog telephone network ( Plain Old Telephone System, POTS ) should form the backbone network. Therefore, the complete ATM standard consists of definitions for the Layer 1 to 3 ( physical layer, data link layer and the network layer) of the OSI model. In charge of the development of ATM standards were mainly telecommunications companies, but also the U.S. Department of Defense ( DoD). For this reason, many of the existing telecommunications technologies and conventions have been integrated into ATM.

Today, the ATM technology to support applications of the global Internet and telephony backbone of the DSL technology is used to the private LAN. The specifications are developed by the ATM Forum. The specifications are then in the ITU-T (formerly CCITT) submitted for standardization.

ATM standards

ATM -layer model ( levels):

Management functions (OAM ) for ATM are defined in a much greater extent than for IP. They include configuration management, fault management and performance measurement.

Tasks of the layers:

The object of the AAL is to adapt to the higher layer data format of the payload field of the ATM cell, and the other side to transmit control information. We distinguish five types of services, of which but so far has only the simplest of greater importance. IP uses the ATM Adaptation Layer 5 ( AAL5 ). The adaptation of the AAL5 mainly handles the fragmentation and reassembly of IP packets that do not fit into the short payload data.

ATM concepts

Why cells?

The reason for the use of small - data "cells" was the reduction of the jitter in the multiplexing of data streams.

Was developed as ATM, were STM -1 lines with 155 Mbit / s ( 135 Mbit / s payload ) fast optical network link, many PDH lines of the former networks were much slower: 1.544 Mbit / s to 45 Mbit / s in the U.S. and 2 to 34 Mbit / s in Europe.

A standard maximum length IP data packet ( 1546 Byte/12368 bit although the IP Specification 64 allows KiB ) needed with these data rates is between about 90 microseconds (135 Mbit / s) and 8 ms ( 1.544 Mbit / s) to transmitted to be blocked at this time and the data channel.

Must now share a split into packets speech signal line ( data channel) with large-volume data traffic, so take these language - no matter how small they are made - always on data packets full size and have to wait sufficiently long to be transmitted can. For voice traffic, these delays were too long, so you would need echo cancellation, even after filtering out the jitter even in local networks. It was at that time simply too expensive.

The solution to this problem was to divide all the packages in 48 -byte packets part, be provided with a routing header of 5 bytes and then to multiplex these 53- byte cells instead of the original packages. Due to the header, the original packets can be later identified and reassembled. This procedure reduced the queuing time by almost a factor of 30, which you could save the echo cancellation.

The rules for sharing and re -assembly of packets and streams in cells are called ATM Adaptation Layers: The two most important are AAL 1 for streams (eg, language) and AAL 5 for almost all kinds of packages. AAL which is actually used is not encoded in the cell. Instead, it is configured between two endpoints or agreed upon on the basis of a virtual connection.

Today, an Ethernet packet full length needs only 1.2 microseconds on an optical link with a 10 Gbit / s data transfer rate, which is actually no longer makes it necessary to use small packets to minimize latency times short. Some conclude that ATM has become superfluous in backbone links.

For slow connections (up to 2 Mbit / s) ATM is still useful. For this reason, many ADSL systems use ATM between the physical layer and a Layer 2 protocol like PPP or Ethernet.

What virtual connections?

ATM is based on compounds that can be both fixed set, as can also be connected via an ISDN -like signaling only for a certain time. Virtual Paths ( VPs) and Virtual Channels were defined (VCs ) for this purpose. Each ATM cell header contains a Virtual Path Identifier ( VPI, 8 or 12 bits) and a Virtual Channel Identifier (VCI, 16 bit). While these cells pass through the ATM network, the switching is achieved by changing the VPI / VCI values ​​. Although the VPI / VCI values ​​therefore, not necessarily from one end of the connection to the other remain the same, this corresponds to the concept of a connection because all packets with the same VPI / VCI values ​​take the same path, in contrast to IP, in which a packet could reach its destination via a different route than previous and subsequent packets.

Virtual connections also have the advantage that you can use them as a multiplexing layer for different services (voice, Frame Relay, IP, SNA, etc.), which then may share a common ATM connection, without interfering with each other.

Traffic management with cells and virtual connections

Another key concept of ATM is called " traffic contract " (Eng. " traffic contract " ): When an ATM connection is set up, each switch is informed by way of the traffic class of the connection.

Traffic Contracts are part of the mechanism is implemented using the Quality of Service ( QoS). There are four basic types (with several variants), each of which describes a set of parameters of the connection:

  • UBR - Unspecified Bit Rate (German indefinite bit rate), the default type for " normal" traffic. Here you can get the bandwidth left over after the QoS traffic is handled. This is therefore a best-effort connection.
  • ABR - Available Bit Rate (Eng. available bit rate ), the transmission rate is controlled according to the current free bandwidth. The control is effected either via the EFCI flag in Cell Header or special resource management (RM ) cells.
  • VBR - Variable Bit Rate ( dt variable bit rate ), here "ordered" to an average cell rate, but you must not exceed a certain amount for a certain time (there are real-time (RT- VBR) and non- real-time variants (NRT - VBR) ).
  • CBR - Constant Bit Rate ( dt constant bit rate ), here is a peak rate Peak Cell Rate ( PCR) is requested, which is then guaranteed. On the other hand, but also that in certain circumstances means Bandwidth remains unused.

Compliance with the Traffic Contracts will normally be enforced by " shaping", a combination of queue and classification of packets, as well as " policing ".

Traffic Shaping

Typically, the shaping happens at the entry point of an ATM network, it is tried there, the cell flow to be controlled so that the traffic contract can be maintained. Simplest form is the Peak Cell Rate (PCR ) shaping which limits the maximum cell rate up to a predetermined value. Shaping within the ATM network requires buffering capacity (Buffer Manager ), because the cells are then delayed occasionally and thus lead to cell cluster.

Traffic Policing

To maintain the network performance, it is possible to provide virtual connections rules against their traffic contracts. If a connection exceeds its traffic contract, the network can either discard the cells themselves or the Cell Loss Priority ( CLP ) bit set to mark the packets for additional switches on the way as discardable. This Policing ie cell works for cell, which can lead to problems if packet-based communication on ATM implemented and the packages were packed into ATM cells. If we reject one of these ATM cells, of course, the whole previously segmented packet is invalid. For this reason they invented schemes such as Partial Packet Discard (PPD) and Early Packet Discard (EPD ), discard a whole series of cells until the next frame starts (see Discard protocol). This reduces the number of redundant cells in the network and saves bandwidth for full frames. EPD and PPD work with AAL5, because they need to evaluate the frame end bit to detect the end of a packet.

Structure of an ATM cell

An ATM cell comprises a header of 5 bytes and 48 bytes of payload (payload ). The payload size of 48 bytes arose as a compromise between the needs of voice and packet-based networks which. You simply took the average of the packet lengths of the American (64 bytes) and the European proposal (32 bytes).

ATM defines two different cell formats: NNI (Network - Network Interface) and UNI (User - Network Interface). Private ATM connections use the UNI format, public ATM networks, the NNI format.

48byte payload

48byte payload

In a cell with UNI header format GFC field for a (to date undefined ) local flow control between the network and user is reserved. Because of this intended use may transfer the GFC bits of public ATM networks is not guaranteed. Until the standardization of the local flow control all four bits must be set by default to zero. In private networks, they can be used in any way, if not manufacturer-specific restrictions prohibit that.

The NNI format of an ATM cell is identical to the missing GFC field with the UNI format. These bits are instead used to from 8 to 12 bits increase the VPI field. Therefore then can a single port 212 VPs, each with 216 VCs addressed and the appropriate number of connections are switched. When UNI format is only 256 VPs, each with 216 VCs. In practice, some VP / VC numbers are usually reserved for special purposes and may not be used for user connections.

The PT field is used to distinguish between different types of cells for user data or maintenance and management purposes. For example: cells for the exchange of signaling information, control data for the control of network elements and cells for resource management and control traffic.

The Cell Loss Priority bit (CLP) indicates whether the cell is a high (CLP = 0) or low priority (CLP = 1) has. This is only important if a node is overloaded and some cells must be discarded. Cells with low priority are discarded first. The CLP bit can be set or changed by terminals or network nodes.

The HEC field (Header Error Correction, checksum of the header ) allows to check whether the header of the ATM cell is corrupted; error checking of user data must be done in higher layers. Furthermore, it serves the cell synchronization: if the receiving side has the cell start not identified correctly, it also takes the wrong bytes as HEC field and then comes long as negative test results until it has synchronized back to the correct cell beginning.

See also: DSS2, DSL, IP, MPLS, DQDB

Numbering in ATM networks

It was originally planned that the discrimination based on ATM technology B- ISDN, the same number space is used as for the ISDN, ie the standardized by ITU-T Recommendation E.164. But after the IT world recognized the ATM technology to be useful, took a violent battle rather than against this numbering scheme. He ended with a number of alternative space was created, which is known under the term "ATM End System Address" ( AESA ) today. It could thus be avoided, that the national telecommunications companies, which often were still monopolies, could dominate the number assignment. Today, both address types are common, but differ in substance:

The former debate about the ATM addressing is similar today to the address in the IP telephony, which has led, among other solutions to the well-known under the name of Telephone Number Mapping.


Almost all operators of communications networks have set up in the backbone area ATM networks, but do not use ATM signaling, but fixed interconnections. In the broadband access network is as a multiplex layer almost exclusively used ATM (DSLAM, RAS). ATM was able to prevail as a technique for local networks only in the high power range. Its high complexity and the associated costs prevented the large-scale use as an integrated network solution in an office environment. Some basic principles of ATM standards such as the ability to prioritize certain types of traffic, were taken later in MPLS, a general protocol for efficient switching below Layer 3

Use in broadcasting and broadcasters

The use of ATM technology is used by send and broadcasters in Germany: over fiber-optic networks to send larger production companies and TV channels in real-time, inter alia, her sound and image material to the various broadcasters - ARD - internal production network ( HYBNET ) based on ATM technology. With the help of satellite technology (uplink) can be even larger distances ( intercontinental) over the ATM network bridge. The transfer points consist of an encoder and a decoder, the muxer so-called ( multiplexed ). The technique also provides the basis for live switching from studio to studio.

Current Situation

For example, although the traditional telecommunications network operators invested huge amounts of money in ATM infrastructure, indications are increasing since the late 1990s that an increasing number of applications, other, often Ethernet-based technologies instead of ATM are used. Reasons for this could be the much lower prices of devices the IEEE 802.3 family and the easier access know-how.

The German Telekom AG plans until 2012 no longer upgrade their Internet broadband connections and their telephone switches with ATM technology, ATM so that in the future does not matter much more in the backbone area. ATM technology is replaced by Ethernet-based technologies and IP-based VPNs. In February 2013 it was announced that Telekom (meaning ATM DSLAMs ) in a test phase to 2016 gradually " little used parts of the conventional network technology " turn off and the customers on IP Solutions ( " All- IP" ) wants to migrate. Affected are several thousand customers who have already been written. The exchanges are to be retained, but the network operator wants to " gain valuable experience for the migration to IP technology, which will then be incorporated into future planning. "

Although ATM is likely to play as a technique that no longer matters, it should be mentioned that some of the research findings with ATM continue to be used in other network technologies, such as in MPLS. But QoS on the Internet or in future networks or the TCP congestion control, has - at least in the field of research - also benefits from ATM ( see TCP / overload control as a research field).