Side_1_360

the total usage charge for best-effort services
is given by:

Usage_chargebest-effort= pV(x)V

where x= {best-effort} and Vis the transferred
volume.

For services with QoS guarantees, a straightfor-
ward approach would be to set the charge per
unit of time equal to pB(x, M), where B(x,M) is
the effective bandwidth for contract xand mean
rate M, and pis the price per unit of effective
bandwidth; the latter price is determined by eco-
nomic factors such as demand and competition).
In this case, the usage charge is then

,

where Tis the duration and Vis the transferred
volume. A disadvantage of such an approach is
that charges are not linear functions of the mea-
surements of duration and volume, thus making
it difficult for users to understand.

Interestingly enough [CA$hMAN, SoKe97],
based on the effective bandwidth bound as a
measure of resource usage, one can construct a
charging scheme linear in measurements of
duration and volume that approximate the previ-
ous charge. This charging scheme is presented
to the users as a trade-off between a duration
charge and a volume charge. In particular, given
his NPL and traffic profile, the user is offered a
set of charging parameters (pT(x), pV(x)) to
choose from. The parameters pT(x), pV(x) repre-
sent a duration and a volume charge, respec-
tively. The usage charge will be

Usage_charge= pT(x)T+ pV(x)V

In practice, for a given SLA, the provider can
offer a small number of tariff pairs.

A number of additions/modifications can be
made to the basic approach described above. If
the service is connection oriented, then one can
include in the usage charge a connection set-up
fee pc[SoKe97]. This fee accounts for the sig-
nalling resources required to set up the connec-
tion and the state that needs to be maintained
throughout the duration of the connection. In
addition, a discount for higher volume connec-
tions can be incorporated in the scheme.

In the case a user generates and injects the traffic
that is not conformant (exceeds the conditions)
with the traffic profile agreed in the SLA be-
tween the user and the provider, various reac-
tions can be initiated. One such reaction is to
mark such traffic for dropping, and charge it
at the same rate as best-effort.

pBx,V

T

⎛

⎝

⎞

⎠

T

7.4 Charging for End Services

Charges for end-services can be given by a for-

mula similar to Equation 1, namely:

Service_chargei= pT(xi)Ti+ pV(xi)Vi+ pc(xi) (3)

where now the parameter xdenotes the service

(application) and quality class selected by the

user. These parameters are included in the SLA

between the user and the SP.

A charge for end-service may include the charge

for the corresponding connectivity necessary to

provide the service, and charges for the service

itself. The latter can include content charges in

the case of e.g. video delivery.

Recall the business model described in Chapter

2, where the SP offers a different quality class to

the user according to the applications he might

use. Hence, the SP “hides” from end users the

low level details of the QoS parameters and their

corresponding values. In the same sense, an SP

may wish to provide very simple tariffs, even if

they lose some, or even all, the structural charac-

teristics of the transport level charges. The moti-

vation for that might be to attract the customers

by having a very simple charging scheme, e.g.

flat-rate charging, or because the transport level

charges are a small percentage of the charges for

the service itself (e.g. the content charge).

Indeed, economic and marketing issues may

have a significant effect on both the structure of

the charging scheme for end services and the

corresponding prices.

As mentioned before, transport level charges at

the SP-NP interface will influence service level

charges. Consider the following examples:

• For an SP offering a video playback service,
  the characteristics and requirements (e.g. in
  terms of bandwidth) for a particular video are
  known in advance. Indeed, these requirements
  depend not only on the content but also on the
  resolution of the video encoding. Knowledge
  of the requirements will in turn enable an SP
  to estimate the corresponding charges of the
  transport level services required by the net-
  work provider for the delivery of the particu-
  lar video. Of course, different video streams
  may have different transport level require-
  ments. The service provider, however, may
  select to offer simple duration-based charges
  with the same prices for all video streams
  which, when averaged over all connections of
  many users, will absorb the varying transport
  level charges.


• For an SP offering IP telephony or videocon-
  ferencing services, the characteristics and
  requirements for a particular session are not