Side_1_360

are harmonised and allow for mapping into

others. Monitoring a QoS parameter could be

done both “continuously” and by “sampling”

according to “typical” usage. The dynamics

of the QoS may have major influence on how

QoS is perceived.

7 Concluding Remarks

One intention is that introducing “sophisticated”
mechanisms related to QoS, and TE in general,
allows the provision of a wider portfolio of ser-
vices and accompanying QoS levels. At the
same time the network resources are efficiently
utilised. These mechanisms introduce a cost in
the form of increased overhead/complexity,
meaning that the basic question is that such cost
must be weighed against the benefits that can be
obtained.

The quality-efficiency product has been pro-
posed by [Bern00], see Figure 10, showing the
trade-off between network utilisation and service
provision guarantees. The illustration gives that
higher loads on the resources, e.g. links, can be
used if less strict values of the QoS are given.
However, introducing more QoS-related mecha-
nisms may allow for higher levels of resource
utilisation for the same level of guarantee.
Hence, if an operator wants to operate a network
efficiently while still supporting strict guaran-
tees, more sophisticated QoS-related mecha-
nisms must be introduced. The different QoS-
related mechanisms may imply different levels
of overhead in terms of processing and storing.

The quality-efficiency product is valid for a cer-
tain network domain. In an end-to-end view,
such a product/measure may not have the same
value for all domains. That is, some domains
may have high utilisation, while for others a low
utilisation is allowed. Naturally, there is no clear
boundary between high and low utilisation, as
between high and low levels of guarantee.

Returning to the issue of demand, not all traffic
flows (and users/applications) will ask for strict
guarantees. One question is whether to define a
number of virtual networks, e.g. with different
quality-efficiency products, on the same physical
network. This would then open for a broader
spectre of service levels, better matching the dif-
ferent customer groups.

In this paper, the basic activities and mecha-
nisms of TE have been described. A motivation
is to provide an introduction to the topics. Many
of the issues are treated in accompanying papers
in this issue of Telektronikk.

Although there are quite a few results available,
essential issues also remain to fully support the
multi-service and multi-provider configuration.

A few of these are briefly described above.

Hence, the continuing need for improving

Traffic Engineering solutions in IP-based net-

works should be beyond doubt.

References

[Bern00] Bernet, Y. 2000. The Role of the Host

Supporting the Full Service QoS Enabled Net-

work. In: Internet2 workshop.Houston. URL:

http://www.internet2.edu/qos/houston2000/pro-

ceedings/Bernet/20000210-QoS2000-Bernet.pdf

[Come88] Comer, D. 1988. Internetworking with

TCP/IP.Englewood Cliffs, NJ, Prentice-Hall.

[ID_IPsm] Eder, M, Chaskar, H, Nag, S. IP Ser-

vice Management in the QoS Network.draft-irtf-

smrg-ipsm-00.txt. July 2001. Work in progress.

[ID_resreq] Kirstaedter, A, Autenrieth, A. An

Extended QoS Architecture Supporting Differen-

tiated Resilience Requirements of IP Services.

draft-kirstaedter-extqosarch-00.txt. July, 2000.

Work in progress.

[ID_tepri] Awduche, D O et al. Overview and

Principles of Internet Traffic Engineering.draft-

ietf-tewg-principles-00.txt. Aug. 2001. Work in

progress.

[Jens01] Jensen, T. Internet Protocol and Trans-

port Protocols. Telektronikk, 97 (2/3), 20–38,

2001. (This issue.)

[RFC2990] IETF. 2000. Huston, G. Next Steps

for the IP QoS Architecture.(RFC 2990.)

Figure 10 Quality – efficiency

product, from [Bern00]

Level of service

provision guarantee

more QoS-related

mechanisms

resource usage efficiency