nomic issues such as market structure and
demand.
Equation (2) is rather general and can describe a
wide range of the charging schemes that are pre-
sent in today’s telecommunications market, for
both guaranteed services and best-effort ser-
vices. Equation (1) can be further generalised by
including time-of-day pricing, which allows
prices to depend on the specific time-of-day the
service is delivered. The rationale for such a
charging scheme is to provide incentives for
users to move traffic which they value less to
off-peak hours (for which the prices are lower
compared to on-peak hours), hence achieving a
more uniform utilisation of network resources
throughout the day.
7.1 Charging for Network
Connectivity Services
In addition to recovering the costs for service
provisioning and generating revenue, charging
may play an important role for controlling re-
source usage. When demand is always less than
supply, the controlling function of charging is
not so important. On the contrary, if demand
exceeds supply, charging can be used as an
effective mechanism for controlling how re-
sources are used, and help the network achieve
efficient and stable operation. In such cases, in
order to provide incentives for users to use the
network according to their actual needs and to
charge them in a fair way, charges need to take
some account of resource usage. Furthermore,
by setting prices appropriately, usage-based
charging can generate the amount of revenue
necessary for expanding the network to meet
the excess demand.
The issues of measurement methods for the
resource usage in networks supporting bursty
traffic and different ways of constructing charg-
ing schemes using these measurements, are dis-
cussed next. Note that the discussion refers to
the usage component of a user’s charge, which
is associated with resource consumption. The
schemes discussed here are appropriate for
charging wholesale transport services, e.g. a
service provided by NP to SP.
7.2 Measuring Resource Usage
The amount of resources used by a user generat-
ing bursty traffic can depend on the user’s traffic
profile, the NPL guaranteed by the network, and
the statistical characteristics of the user’s traffic.
It is desirable to map such multi-dimensional
quantities into a scalar that reflects the relative
amount of resources used by the user. This
scalar is typically called the “effective rate” or
“effective bandwidth” of the user’s traffic
stream, and can simplify the problem of charg-
ing a network service based on the relative
amount of resources used by the service.In the case of best-effort services, the effective
bandwidth of a stream reduces to its mean rate.
This can be understood as follows: For best-
effort services, there are no performance guaran-
tees. The only requirement is that traffic eventu-
ally reaches its destination. Considering a single
link, the latter requirement translates to a stabil-
ity condition for the link that can be written as,
where miis the mean rate of a traffic stream i
and Cis the link capacity. Hence, the mean rate
is the appropriate measure of resource usage for
best-effort services.In the case of guaranteed services, the actual
effective bandwidth of a traffic stream is a com-
plex function, and one usually considers bounds
of the actual effective bandwidth that are simpler
and involve easy to measure quantities. In the
case where the only measure of resource usage
is the mean rate, the bound can be written as
B(x,m), where xincludes the NPL and traffic
profile, and mis the mean rate of stream. It can
be shown that this bound is a concave function
of the mean rate m[CA$hMAN].The concavity of the effective bandwidth bound
is large when the peak rate is high relative to the
network capacity or when the QoS guarantees
are tight (e.g. small delay and small loss proba-
bility). This concavity property can be used to
provide interesting incentives to the users. On
the other hand, for best-effort services the curve
becomes linear, i.e. the effective bandwidth is
equal to the mean rate of the stream, as dis-
cussed above.More complex bounds that depend on more
detailed statistics than the mean rate would
result in higher accuracy. However, investiga-
tions and trials have shown that higher complex-
ity, hence greater difficulty to understand charg-
ing schemes based on such bounds, can easily
outweigh the advantage of higher accuracy
[CA$hMAN].7.3 Charging for Resource Usage
The approaches, discussed before, for measuring
resource usage provide input to the charging
scheme. For best-effort services, it was men-
tioned that the appropriate measure of resource
usage is the streams mean rate. Hence, the
charge per unit of time is p(x)m, with x= {best-
effort}. The parameter p(x) is the price per unit
of rate and unit of time for best effort services;
this price will typically depend on economic fac-
tors such as demand and competition. Hence,