Computersight -

ATM Distinguished Service Record

Traditionally ATM has a long and distinguished service record for voice communications. It is also ideally suited to multiplexing environments and can be readily configured to carry VoIP traffic streams.

In fact today we find that most consumer ADSL2+ implementations do offer a choice of PPPoE or PPPoA as their transport protocols (at least here in Perth). PPPoA stands for Point-to-Point Protocol over Asynchronous Transfer Mode.

The importance of this cannot be overlooked as it means that ATM in some form or other will be with us for some time into the future. In fact the Japanese have just recently deployed a communications satellite with an onboard ATM switch. They obviously think there is life in ATM yet.

Introducing Asynchronous Transfer Mode (ATM)

Originally intended to be a unified networking strategy Asynchronous Transfer Mode (ATM) is a connection-oriented, circuit-switched, cell relay “Jack-of-all-trades” transport protocol that uses small uniform fixed-sized cells to redress Quality of Service (QoS) issues so important to voice/video communications and the multitude of streaming applications upon which we are all so dependant.

ATM Origins and Development

During development of the standards for the Asynchronous Transfer Mode (ATM), in the mid 1980s, the goals were to create a unified networking strategy that could act as an all-round transport system for real-time video and audio as well as for image, text and email. ATM is pretty much a “Jack-of-all-trades” transport system. The two groups primarily responsible for the development of the ATM standards were the International Telecommunications Union [ITU 2004] and the ATM Forum [ATM 2004].

Over time we have seen that the majority of implementations and uses that ATM has fulfilled have been primary concerned with telephony and IP networks. Ethernet and the Internet Protocol (IP) are packet-switched network technologies that use packets of variable size referred to as frames.

ATM Protocol Basics

In marked contrast to packet-switched networking technologies; ATM is a connection-oriented, Data Link Layer (OSI Reference Model Layer 2), circuit-switched, cell relay protocol that runs over Synchronous Optical Network (SONET) Physical Layer links (OSI Reference Model Layer 1) using cells of identical and never varying size. Consistent predictability is the underlying ethos here.

Being a connection-oriented channel-based technology means that ATM must always establish a “logical” connection between the two endpoints prior to commencement of data exchange. Significantly, ATM encodes data traffic into small uniform fixed-sized cells. ATM cells are always 53 bytes in size and are comprised of 48 bytes of data and 5 bytes of header information.

ATM Cell Structure

Regardless of the original size of the packets to be transmitted ATM breaks all packets, data, and voice streams into 48-byte chunks and then adds a 5-byte routing header to each one thereby making a total of 53-bytes for each and every cell. The 5-byte header is essential for later reassembly. During development of ATM it was considered that 10% (5 bytes) of each cell (payload) being dedicated to the header for routing information was more than sufficient.

ATM multiplexes these 53-byte cells instead of the larger packets and in so doing reduces the worst-case queuing jitter by a factor of almost 30, thereby removing the need for echo cancellers. I will discuss queuing jitter along with other types of jitter shortly.

ATM Cell Formats

ATM defines two different cell formats the Network-Network Interface (NNI) and the User-Network Interface (UNI). Most ATM links use the UNI cell format.

ATM Adaption Layers (AAL)

ATM Adaptation Layers (AAL) are the rules for segmenting and reassembling packets and streams into cells. It is the AALs that provide the support for the various services delivered by ATM.

Currently, there are five different AALs and the information concerning which one is being used for each cell on a cell-by-cell basis is not contained within the cell or in the cell header. Rather, this information is negotiated by or configured at the endpoints on a per-virtual-connection basis. Here are the five different AALs and their main uses:

1. AAL1

   Constant Bit Rate (CBR) Services, Circuit Emulation

2. AAL2

   Variable Bit Rate (VBR) Services

3. AAL3

   Variable Bit Rate (VBR) Services

4. AAL4

   Variable Bit Rate (VBR) Services

5. AAL5

   Data Transport

ATM Connectivity

Because ATM is a connection-oriented channel-based technology it must establish a “logical” connection between the two endpoints prior to commencement of data exchange. ATM does this by implementing Virtual Circuits, Channels, Paths and Identifiers as follows:

• Virtual Circuits (VC)

  Virtual Circuits (VC) are admirably suited to multiplexing scenarios. Simply by including an 8-bit or 12-bit Virtual Path Identifier (VPI) and a 16-bit Virtual Channel Identifier (VCI) pair in every ATM frame’s header each Virtual Circuit (VC) is uniquely identifiable.

• Virtual Channel

  An ATM Virtual Channel represents the basic means of communication between two end-points. Cells are given a unique identifier called the Virtual Channel Identifier (VCI) which is placed into the ATM cells’ header. All ATM cells containing identical VCIs are transported in the same Virtual Channel.

• Virtual Path (VP)

  A Virtual Path (VP) denotes the transport of ATM cells belonging to virtual channels which share a common identifier called a Virtual Path Identifier (VPI). The VPI is included in the header of every ATM frame. In other words a Virtual Path (VP) is a bunch of Virtual Channels (VC) connecting the same end-points. These will also have a common traffic allocation.

• Virtual Path Identifier (VPI)

  The Virtual Path Identifier’s (VPI) length varies depending on the interface it is sent on (inside the network or on the edge of the network.

ATM Traffic Contracts

When an ATM circuit is set up each ATM switch is informed of the traffic class of the connection. These ATM contracts constitute part of ATM’s Quality of Service (QoS) mechanisms. There are four basic types of contracts:

1. Constant Bit Rate (CBR)

   A constant specified Peak Cell Rate (PCR) is set

2. Variable Bit Rate (VBR)

   An average cell rate is specified. This may peak at a certain predefined maximum level for a certain length of time before becoming problematic

3. Available Bit Rate (ABR)

   A minimum guaranteed rate is specified

4. Unspecified Bit Rate (UBR)

   Traffic is allocated all remaining transmission capacity

ATM Traffic Contract Delivery and Monitoring

Traffic Shaping

The intended objective of traffic shaping is to ensure that cell flow will meet its traffic contract and is usually done at the entry point to an ATM network.

Traffic Policing

To maintain network performance it is possible to “police” virtual circuits against their traffic contracts. Basic policing works on a cell by cell basis, but this is sub-optimal for encapsulated packet traffic. If a circuit is exceeding its traffic contract, the network can either drop the cells or mark the Cell Loss Priority (CLP) bit (to identify a cell as being discardable farther down the line).

Benefits of Using Small Fixed Size Cells

The major benefits derived from using small data cells are a reduction in queue delay and jitter; particularly in multiplexing data streams. By using small, fixed-sized cells ATM is able to transport large data files all the while maintaining minimal queuing delays. Minimal queuing delays are essential to the delivery of both voice/video communications.

Queue Delay

Queue delay related issues include problems associated with end-to-end-round-trip delays and delay variance particularly when carrying voice traffic. High traffic volumes and/or congested networks along with the arrival variance associated with variable route routing are among the main causes of queue delay issues.

Jitter

Although jitter results from queuing delay issues deviations or displacement of various aspects of high frequency pulses such as amplitude, phase timing and signal pulse width as a direct result of electromagnetic interference (EMI) and crosstalk (noise) also cause jitter. Think of jitter as being the production of “jerky” results or in video applications flicker. By using small fixed-size cells ATM is able to overcome the effects of queue delay as well as other types/sources of jitter.

Multi Purpose Transport Protocol

Asynchronous Transfer Mode (ATM) carries many different data types and formats (text, audio, video, graphics, photos etc.) from a multitude of sources and of variable sizes. When combined with standard queuing strategies, maximum queuing delays were common. This is totally unacceptable where voice and real-time video traffic is concerned.

Compression/Decompression Algorithms (Codec)

Because of the way in which many Compression/Decompression Algorithmswork special considerations need to be implemented in order to ensure they work properly as intended including:

Time

The nature of time as we humans perceive it is an analogue continuum (that is to say time is a linear progression). Once past, there is no way as yet to recover the loss.

Jitter and Queue Delay

Jitter and queue delay are of great importance because of the nature and manner of operation of the compression/decompression (codec) algorithms used in the conversion of a digitalized data stream back into an analogue audio signal. This conversion process (digital-to-analogue) is very much a “real-time, on-the-fly” process and is more attuned to” just-in-time” transport protocols.

Real-Time Streaming

In order to produce reliable, consistently “acceptable” output the codec needs the data items (the digitized voice data) to be presented to it in a predictable, regulated and evenly spaced in time data stream, hence the term “real-time streaming”.

Late Arrivals

If the data arrives after its allotted position/reception window in the time sequence (relating to that part of the data-stream) the codec will simply drop it. Not surprisingly this is unacceptable for IP telephony. Remember to keep in mind that time is analogue in nature and once a “time window” elapses, the “lost” time becomes unrecoverable.

Codec Packet Handling Options

If the transport protocol is unable to present the data as and when the codec expects it, the codec, has no choice but to assume either silence, make a “best guess” or simply drop the packet. Any way is unacceptable where voice is concerned as the conversation rapidly becomes untenable and the message does not get through.

ATM Deployment Indicators and Scenarios

ATM WAN Core Implementation

ATM production environment implementations have over time proved to be very successful in the Wide Area Network (WAN) scenarios. Numerous telecommunication providers and Internet Service Providers (ISPs) have implemented ATM in their Wide Area Network (WAN) cores.

Slow Links

For slow links less than 2M-bit/s, ATM still makes sense, which is why many ADSL systems use ATM as an intermediate layer between the physical link layer and a Layer 2 protocol like PPP or Ethernet.

Linear Audio and Video Streams

Interest in using native ATM for carrying live video and audio has increased recently. It is in these environments, where ATM can deliver the low latency and very high Quality of Service (QoS) required for handling linear audio and video streams.

Gigabit Ethernet

Today we are finding that for both new WAN implementations and for existing WAN implementation upgrades, high speed, high performance Ethernet (Gigabit Ethernet, 10Gbit Ethernet, and Metro Ethernet etc.) are rapidly replacing ATM as the technology of choice.

Relative Performance

At the time ATM was designed, 155Mbit/s (135Mbit/s payload) over fiber-optic cable was very fast in comparison to the other carrier/transport technologies available at the time. Since then however; these other technologies have evolved and are now considerably faster than they once were.

Jitter

Today; a 1,500 byte (12,000 bit) full-size Ethernet packet takes only 1.2 µs to transmit across a 10Gbit/s optical network. With this sort of speed, jitter is no longer the issue it once was. By overcoming the potential adverse effects of jitter through this ramping up of network transfer speeds we have at the same time removed the need for using small uniform cells to overcome jitter.

Complexity

Unfortunately, due to ATM’s complexity it proved to be unsuitable for deployment in many of the scenarios that its creators had originally intended.

Converged Networks

The speed and traffic shaping requirements of many converged networks are also proving to be very challenging for ATM.