The cost of Internet service (bandwidth) is skyrocketing and at the same time user demand for bandwidth is also exploding. The problem we face, as managers of network service for our institutions, is how to explain to management what we're paying for and to users why they are going to have to pay more.
Internally we also need to understand how to evolve our network infrastructures most cost effectively in order to meet our users' increasing expectations.
These problems argue for well understood metrics and models for characterizing our network behavior and the impact of different types of applications on that behavior. Current measures of offered and delivered load may reflect as much about the network infrastructure as about the impact of types of applications. It is almost impossible therefore to use this data to predict utilization growth or justify increasing service costs.
Clearly the problem of identifying service (what we're paying for) must be solved cooperatively and collaboratively between service providers and network consumers. At the very least we must be able to measure the same things so that we can achieve independent verification of the service characteristics. Better yet, we must be able to analyze the measures in terms of realistic models of network behavior so that we can work together to optimize our service interface and infrastructure requirements.
As an example, consider the current model of charging for offered load. All my ISP has to do is drop half my packets somewhere beyond my gateway router and they can increase the amount they will charge me by 50%!! Obviously this simple measure of "service" is not only inadequate but grossly distorted.