Side_1_360

served, longer buffer sizes may apply as these

would accept longer delays (and delay varia-

tions), although the loss requirements might not

be lower.

One of the primary goals of TE in an operational

network is to limit the sustained congestion.

A focused congestion may result from unbal-

anced mapping of traffic flows onto the resource

groups. Then, means can be activated to dis-

tribute the traffic flow in a better way. TE mech-

anisms do also allow for differentiating and

ensuring service levels. This would meet the

characteristics related to the spectrum of traffic

flows. An example is to use separate real-time

and non-real-time traffic on the buffering side,

while still using the same link capacity. Hence,

high link utilisation is achieved, while knowl-

edge of the traffic flow characteristics is ex-

ploited. Delay, delay variation and loss ratio are

examples of characteristics used so far. Another

important parameter is related to dependability,

such as the availability of the service. This

would also be addressed by the TE-related

mechanisms, increasing the general depend-

ability but also allowing for differentiation.

In sum, the TE mechanisms offer a set of “tools”

that a network operator can tune for its opera-

tion, reaching better utilisation of network

resources while allowing predictable service

levels (differentiated and ensured).

The actual need for introducing TE-related

mechanisms is questioned by some. Two factors

supporting this view are the traffic growth and

the willingness to pay. For the former, it is con-

sidered that by the traffic growth seen these days

there is little need for accurate procedures as

more capacity should be installed all the time.

Hence, additional capacity present when over-

provisioning would fairly soon be needed any-

way. Regarding the latter factor, much traffic on

IP-based networks comes from Web browsing,

assumed to be looking for low-priced, low ser-

vice level.

An argument in the other direction is that as one

of the current trends is that more commercial

activities are based on IP networks, a stronger

demand for predictable services will emerge.

This also seems to be recognised by the industry

in this area as much work is allocated to TE-

related mechanisms.

In this article, some of the most central themes

of TE are described. In the following chapter

overall objectives, scope and resource types are

described. Chapter 3 presents the activities,

processes, key components and mechanisms

together with the contexts in which the TE activ-

ities are carried out. Some requirements on TE

systems are listed in Chapter 4. Chapter 5 ex-

plains the TE taxonomy, while Chapter 6 elabo-

rates on further challenges in the TE and QoS

areas.

2 TE Objectives and

Resource Types

The main goals of traffic engineering are to

improve performance for IP-based traffic while

still utilising the network resources efficiently.

In the TE framework principles, as elaborated

within IETF, architectures and methodologies

for evaluating and optimising the performance

of operational IP networks are addressed. The

framework as described in [ID_tepri] gives both

the terminology (set of key terms) and the taxon-

omy (criteria for describing a system). Internet

traffic engineering is here defined as

that aspect of Internet network engineering

dealing with the issue of performance evalua-

tions and performance optimisation of opera-

tional IP networks. Hence, measurement,

characterisation, modelling and control of

traffic are included.

A major objective is to improve the performance

of operational networks, at the traffic and at the

resource levels. This is striven for by looking at

traffic-related performance requirements, like

delay, delay variation, packet loss and goodput.

At the same time the network resources should

be efficiently utilised. An essential part is to

achieve reliable network operations, e.g. during

failures. Having efficient routing configurations

is also central as these decide the way the pack-

ets are distributed in the network.

Capacity management and traffic management

can be used for the optimisation done as part of

TE. Capacity arrangement includes capacity

planning, routing control and resource manage-

ment. The resources include link bandwidth,

buffer spaceand computational, see Figure 3.

Figure 2 Schematic illustra-
tion of relations between delay
and loss

loss

high

low

increasing

traffic load

increasing

buffer size

short long

delay