Side_1_360

to be cross-connected for the cases with varying
weight factor. The effective bandwidth has a sig-
nificant dependence on the LSP capacity, and in
the case of decreasing the demand the effective
bandwidth has a higher relative increase. There-
fore the case with reduced demand is more sen-
sitive to the number of LSPs.

For all cases the sum of the required bandwidth
on the LSPs on a link is less than the actual link
bandwidth, and the link bandwidth not allocated
is illustrated in Figure 16 a). When the number
of LSPs is increased from the minimum number
to the maximum number of LSPs, the required
total LSP bandwidth increases. The additional
bandwidth caused by establishing LSPs is
depicted in Figure 16 b). In these examples, the
last type of traffic flow in Table 2 (i.e. Flow
id 5.2) has been assigned its mean rate during
the capacity calculations. This is to see the effect
of less guarantees to the Telegame application.
The superfluous capacities on links could be
used for additional traffic loads, both giving
higher rates to ongoing sessions and potentially
accepting more sections (when admission con-
trol is used).

8 Measuring Traffic and

Performance

8.1 IP Performance Metrics

In order to reach a situation where users and
providers of IP services have a harmonised
understanding of performance of the network,
a set of harmonised IP performance metrics has
been devised. The following criteria have been
identified to achieve a common understanding,
ref. [RFC2330]:

• The metrics must be concrete and well-
  defined.


• A methodology for a metric should have the
  property that it is repeatable (i.e. same results
  from applying the method several times under
  identical conditions).


• The metrics must exhibit no bias when identi-
  cal technology has been used to implement the
  IP network.


• The metrics must exhibit understood and fair
  bias for IP networks implemented with non-
  identical technology.


• The metrics must be useful to users and
  providers in understanding the performance.


• The metrics must avoid inducing artificial
  performance goals.

Figure 15 Results from a selection of example variations: a) Number of LSPs;

b) Total, k = 1.0; c) Relative cost for minimum number of LSPs and maximum

number of LSPs related to the lowest obtainable cost

Figure 16 Available bandwidth: a) Link bandwidth not allocated to LSPs;

b) additional bandwidth caused by establishing LSPs

reference case single high demand one tenth demand

min LSPs max LSPs

c)

Relative cost

1.15

1.10

1.05

1.00

0.95

850

650

450

250

50

0.01 1 100

Weight factor, k

a)

Number of LSPs

high demand

reference

one tenth demandsingle CoS

Million

200

100

0

Total cost

b)

reference

case singlehigh

demandone tenthdemand

single CoS

high demand

25000

20000

15000

10000

5000

0

0.01

Weight factorr, k

b)

Additional bandwidth

Mbit/s

1 100 10000 1000000

reference

one tenth demand

reference

single CoS

high demand

5000

4000

3000

2000

1000

0

0.01

Weight factor, k

a)

Not allocated bandwidth

one tenth demand

Mbit/s

1 100 10000 1000000