Unit 2

Lesson 8 - ATM Protocols

   Transfer Modes

A transfer mode specifies a method of transmitting, multiplexing, and switching data in a network. Three transfer modes were considered as possible candidates for B-ISDN, including:

Synchronous transfer mode (STM)

Packet transfer mode (PTM)

Asynchronous transfer mode (ATM)

Synchronous Transfer Mode (STM)

"Synchronous" means that data communication is organized by a microprocessor clock; a receiving node can detect the beginning and end of a signal because signals start and stop at particular times. Networks that use STM divide each transmission frame into a series of time slots, and then allocate particular time slots to each user. For example, on the Synchronous Transfer Mode Diagram, Time Slot 2 is dedicated to the same user in each and every frame.

Synchronous Transfer Mode

Synchronous Transfer Mode

STM is ideal for transmission of voice and video, because it provides a constant-bit rate service. Voice and video require predictable and guaranteed network access, or the quality of the transmission degrades rapidly.

In contrast, data transmissions are typically bursty; a user is idle for relatively long periods of time between short periods of intense data-transfer activity. For example, in a typical client-server transaction, the client request consumes very little bandwidth. However, when the server responds, a large amount of data is typically transmitted from the server back to the client. The server's response may actually consume the entire bandwidth of the network.

STM is inefficient for data communications because the same time slot in each frame is reserved for a particular user, regardless of whether the user has data to transmit. When a user is idle, the time slot is wasted, because STM does not reassign unused time slots to other users.

Examples of STM technologies include standard T1 circuits, E1 circuits, and Synchronous Digital Hierarchy (SDH) circuits provided by telecommunications carriers.

Packet Transfer Mode (PTM)

In a network technology based on PTM, data is broken into variable-size units of data (packets, datagrams, or frames). Each unit contains both user data and a header that provides information for routing, flow control, and error correction. Instead of establishing a dedicated physical connection between the source and destination station, the network relays packets from one node to another, often in multiple parallel paths, until they reach their final destination. PTM is implemented through technologies such as Ethernet, Token Ring, Fiber Distributed Data Interface (FDDI), X.25, and frame relay. The Packet Transfer Mode Diagram illustrates the concept of PTM.

Packet Transfer Mode

Packet Transfer Mode

PTM is excellent for bursty data applications because a station only consumes bandwidth when it needs to transmit data. When a station is idle, its share of network bandwidth can be used by other stations. The resulting variable transmission rates and delay, within reasonable limits, are not critical issues for data communications.

However, PTM does not provide the guaranteed network access required by constant-bit rate applications such as voice or video. Voice and video tolerate very little delay in transmission; however, they can handle some loss or inaccurate information.

Asynchronous Transfer Mode (ASTM)

Thus far, we have seen that STM is excellent for voice and video applications, but it is inefficient for data applications. On the other hand, PTM is excellent for data applications, but cannot provide the guaranteed bandwidth and low delay required for voice and video.

ATM offers the best of both worlds. It combines the strengths of STM (constant transmission delay and guaranteed capacity) and PTM (flexibility and ability to handle intermittent traffic) in a single transfer mode that meets the needs of voice, video, and data applications.

In computing, the term "asynchronous" usually means that data transmission is coordinated through start and stop signals, without the use of a common clock. However ATM networks use "asynchronous" to describe how network bandwidth is assigned to user applications. ATM assigns network access to users based on demand, which means that locations in the synchronous data stream are assigned to users in a random, or asynchronous, pattern. The Asynchronous Transfer Mode Diagram illustrates the concept of ATM.

Asynchronous Transfer Mode

Asynchronous Transfer Mode

Fixed-Length Data Cells - ATM organizes transmission by formatting data into fixed-size units called cells. Each cell contains 53 bytes that are divided into a 48-byte payload (data) field and a 5-byte header. The fixed length makes it simpler and faster for an ATM switch to process cells. ATM's cell-switching approach also makes efficient use of network bandwidth for bursty data transfers, by allocating cells to applications only as needed. The Asynchronous Transfer Mode Cell Diagram illustrates the basic format of an ATM cell.

Asynchronous Transfer Mode Cell

Asynchronous Transfer Mode Cell

Connection-Oriented Transmission - In an ATM network, a pair of source and destination nodes establishes a virtual connection before the source begins transmitting data. All cells transmitted between a pair of source and destination nodes follow the same virtual connection or VP (through a network of ATM switches) during the transmission. A later transmission between the same source and destination may follow a different VP, but the path will not change for the duration of the transmission. This approach improves overall transfer speed by making it simpler and faster to switch cells through intermediate nodes.

The ultimate goal of ATM is to provide extremely high-speed communications that allow voice, video, and data applications to run across a single integrated network. By combining connection-oriented transmission with the use of small, fixed-length cells, ATM solves many of the problems encountered when these applications share the same network.

The connection-oriented nature of ATM provides minimal delays for voice and video applications.

The use of fixed-length cells simplifies switch design by allowing the switch logic to be implemented in silicon, in other words, in the switch firmware instead of in software. This greatly reduces the processing time required for each cell, increases switch throughput, and reduces the cost of the switching technology.

Small video cells are not delayed by large data cells because all cells are the same size. This means it is relatively easy to predict the amount of network delay between any two points. In addition, the variation in delay is significantly decreased, because time-sensitive applications such as voice and video can share the same transmission facilities with data applications.

The use of fixed-length cells rather than time slots overcomes the major weakness of STM-wasted bandwidth. An ATM station only consumes bandwidth when it has data to transmit.

[ Previous Section ] [ Unit Contents ] [ Next Section ]