Teletraffic engineering in broadband networks

Teletraffic engineering is a well-understood discipline in the traditional voice network, where traffic patterns are established, growth rates can be predicted, and vast amounts of detailed historical data are available for analysis. However, for modern broadband networks, the teletraffic engineering methodologies used for voice networks no longer suffice [#REF| [1] . Various aspects relating to teletraffic engineering in broadband networks are discussed in this article.

Firstly, the nature of broadband traffic is different from that of traditional voice networks. Many of the methodologies developed for traditional networks were based on the nature of voice calls, and are therefore not applicable to broadband networks. The nature of broadband traffic (broadband traffic characteristics) is discussed in the following sub-section.

The inherent nature of broadband networks is also different from that of traditional voice networks. Broadband networks have:

* high speeds,
* small cell sizes (in ATM networks), and
* limited information in the header.

These factors make teletraffic engineering in broadband networks more difficult than in traditional voice networks. A few more factors that further complicate teletraffic engineering in broadband networks [#REF| [2] are:

* A wide range of applications with diverse Quality of Service (QoS) requirements must be catered for.
* Much of the traffic (e.g., voice, video) is not amendable to flow control.
* The feedback within the network is “slow”.
* There are a large variety of traffic patterns (see long-tail traffic).

Broadband traffic characteristics

In the traditional voice network, the study of traffic characteristics has matured over many years following the seminal work of Erlang in 1909. However, the teletraffic theory that has evolved relied heavily on the facts that

* Only one type of connection is offered, a circuit-switched connection, and
* the order of magnitude of the call holding time is relatively stable and well-known, namely the few minutes of a telephone conversation.

The diversity of broadband service connections (see broadband networks) and the variety of holding times make the application of teletraffic theory in voice networks to broadband networks very difficult [#REF| [3] . Figure 1 and Figure 2 show some applications and the variation in holding time and burstiness that may be expected for each one.

To manage the traffic implications of all these types of connections, we must return to the basic principles of traffic statistics. This has been extensively studied in recent years, and there is a large volume of published work on the subject. A Poisson process with one parameter does accurately model telephony traffic. However, to account for the changes with broadband traffic, alternatives to the Erlang formula have been considered [#REF| [4] . Two methods of modelling the traffic are considered, namely the Bernoulli-Poisson-Pascal approximation and the Maximum Entropy method. These methods use two parameters for describing the traffic – one for the mean demand and another to characterize the variability of the traffic. (See also Long-tail traffic).

Mechanisms used for teletraffic engineering in broadband networks

There are two primary mechanisms that are used for teletraffic engineering in broadband networks, namely:

* Traffic Control and Management, and
* Congestion Control

These two mechanisms are described in the following two sub-sections.

Traffic control and Management

"Traffic control and management" is defined as [#REF| [2] the set of actions performed by the network to:

* avoid congestion
* ensure that users get their required Quality of Service (QoS) guarantees.

The basic control problem is related to the efficient allocation of network resources so as to:

* satisfy different QoS requirements; and
* provide fair access to the network resources for all users

Traffic control and management provides the means by which:

* a user is ensured that the offered cell flows meet the rate specified in the traffic contract; and
* the network is ensured that the traffic contract rates are enforced such that the Quality of Service guarantees are provided for all users.

Congestion Control

"Congestion control" is defined as [#REF| [2] the set of actions performed by the network to prevent or reduce congestion. Congestion control is the most important part of the traffic management issue.

A network that controls congestion must:

* be responsive to the different utility functions of the users
* be able to manage the resources so that there is no loss of utility as the load increases

See Congestion Control for the full article.

A simple example

The process of how a broadband network provides different levels of service to different types of traffic while avoiding congestion will be described briefly by means of a simple example. Links are provided for the reader interested in a more detailed description of the various concepts involved.

The concept of traffic control in broadband networks (particularly in ATM networks) is very simple: an application that requires the network to transport traffic from one location to another with a specific Quality of Service (QoS) follows the following procedure [#REF| [5] :

* The application declares the traffic’s characteristics and the Quality of Service required by the traffic in a traffic contract (in order to make a connection request).
* The network judges whether it has enough resources available to accept the connection, and then either accepts or rejects the connection request. This is known as admission control.
* If there are insufficient resources available in the network at the time, the connection request is rejected.
* If there are sufficient resources available in the network at the time, the connection request is accepted and the network assigns the resources necessary to the connection.
* During communication, the network monitors the conformity between the declared traffic characteristics (in the traffic contract) and the characteristics of the actual traffic entering the network. This is known as traffic policing.
* The network has the capability to discard the non-conformant traffic in the network (using Priority Control).
* The overall rate at which all the traffic is admitted into the network is governed by the network’s traffic shaping policy.

The ATM Traffic Management Framework

The ITU-T has defined a collection of ATM control mechanisms that operate across a spectrum of timing intervals [#REF| [6] . These control mechanisms are summarised in table 1.

ee also

* Broadband Networks
* Teletraffic engineering
* Long-tail traffic
* Traffic Contract
* Network Resource Management
* Connection Admission Control (CAC)
* Usage Parameter Control and Network Parameter Control
* Priority Control
* Traffic Shaping
* Flow Control

=

References

=

[1] Nguyen L. T., Fundamental of Online Traffic Engineering, http://216.239.59.104/search?q=cache:euaQfB_6w8AJ:www.zvolve.com/pdf/LucNguyenUSTA.PDF+fundamentals+of+online+TE&hl=en&client=firefox-a, Last accessed 25 February 2005.

[2] Modern techniques of networking, Traffic and Congestion Control in ATM networks, http://scholar.google.com/scholar?hl=en&lr=&safe=off&q=cache:8j8I9JNI5WQJ:www.itm.hk-r.se/~adrian/courses/modern_techniques_networking/lectures/Lecture_8.ps+teletraffic+engineering+traffic+policing, last accessed 14 March 2005.

[3] Sexton M., Reid A., “Broadband Networking: ATM, SDH and SONET”, Artech House Inc., Boston, London, 1997. ISBN 0-89006-578-0.

[4] Molnar, S., Farago, A., Henk, T., and Blaabjerg, S., Towards precision tools for ATM network design, dimensioning and management, Technical University of Budapest, Hungary, 1995.

[5] Saito, H., Teletraffic Technologies in ATM Networks, Artech House, 1993. ISBN 0-89006-622-1.

[6] ATM Forum Traffic Management Specification, Version 4.0, 1996.