| Abstract: |
The significant increase of internet traffic is to a large extent caused by
more high-data-rate applications like file-sharing etc. Although network
operators constantly increase router and line capacities, overload occurs from
time to time, causing delays, jitter and packet-losses at the data packet
level. At the service level this may significantly reduce the quality of
certain applications. Among those are interactive services like VoIP, some
business applications, online gaming etc. and other on-time-services like
internet-television.<p> This will result in systematic inefficiencies which
can mainly be attributed to two reasons. The first one is the widespread use
of internet-flatrates and the second one is a much too strict interpretation
of the network neutrality principle. The latter describes the fact that every
single data packet will be handled strictly equal at every router, no matter
what application it belongs to and what the technical and economic
consequences of delayed or dropped packets will be. <p> A congestion model
shows that flatrates which users' marginal outlays cause to be zero are
inefficient as soon as positive marginal overload externalities exist. It
comes up with a general pricing solution which will be questioned later on. In
this framework, optimal internet capacity is identified and the networks'
overprovisioning policy is found to be inefficient. <p>The key issue for
internet congestion problem is the fact that, although services are homogenous
at the data packet level, they are very heterogenous at the service level.
They are very different with respect to data rate, quality-sensitivity, and
economic value.<p> Under strict network neutrality rule it can be demonstrated
that certain valuable, quality-sensitive services will be significantly harmed
(and potentially be crowded out altogether) by non-qualitysensitive,
high-data-rate services which may have low economic value. Giving priority to
certain services in overload situations looks like the adequate solution to
the problem. However, this always bears the risk of discriminating some
service providers and applications and will be heavily debated. A more
appropriate solution is provided by priority pricing, whereby users express
their willingness to pay for priority treatment in case of an overload.
Customers have an ex ante choice between different qualities of service. The
choice of a service provider to pay for high priority (high quality of
service) will depend mainly on two factors, the quality-sensitivity and the
end-users' willingness to pay for such services. Only providers of
quality-sensitive services will have any reason whatsoever to pay for traffic
prioritization. Providers of non-quality-sensitive services (file sharing,
e-mailing, webbrowsing) will be adequately served by best effort traffic and
will thus obtain it cheaply. <p> Priority pricing (quality of service) results
in an economically efficient use of scarce router capacity according to the
economic congestion effects of the specific service. It avoids the
crowding-out problem. It allows to generate more economic value out of a given
internet capacity. |