This is a paper about pricing Internet services. More specifically, it is a paper about algorithmic bases to support equitable and efficient charging for such services. The authors state early in this paper that while “differentiated services are demanded by many [Internet] applications, the optimal set of mechanisms to control this differentiation, by balancing economic and technical efficiency, has not been achieved.”

This statement defines the goal of the paper: to delineate such mechanisms. The authors point out the need for incentive-compatible pricing schemes that limit consumption of rare resources, and that provide the requisite degree of feedback to both providers and users of these services. Central to developing pricing strategies (tariffs) is the need to develop ways of measuring usage of different services on the network.

Measuring usage naturally leads to a focus on improving performance, as well as pricing and charging for services. They are all interrelated. One key problem is to figure out how to provide good quality of service (QoS) by differentiating services (and measuring usage of services) in an integrated networking technology architecture. The authors contend that solving this problem has been historically attempted by utilizing technical schemes that rely too much on “bandwidth-driven over-provisioning of network services,” and that this does not really solve the QoS problems.

This paper is dense, and is written in a complex style that clouds what the authors are trying to say. It is grammatically correct to be sure, but it is often difficult to follow. Having said this, the content of the paper makes it worth the extra effort needed to understand what is being said. The basic construct is the cumulus pricing scheme (CPS). The authors contend that this approach simultaneously addresses the issues of network efficiency, user acceptance, and technical feasibility in concrete, workable ways.

CPS is a contractual agreement for service between a user and a provider based upon the long-established idea of a service level agreement (SLA). Users accumulate points (cumulus points (CPs)) that indicate the over- or under-utilization of services, based upon the agreed levels. Over time, these cumulus points provide feedback that can be used to adjust the SLAs. Over time, over an entire network, this approach allows for the fine-tuning of usage, expenditures, and overall network efficiency.

This fine-tuning derives from the application of multiple perspectives in timeframe for the analysis of the SLAs. Each timeframe addresses a different aspect of network management, from applications in the shorter term, to pricing in the medium term, to network capacity management in the longer term, collectively addressing technology and economics in a seamless framework.

Charging dimensions, time-scales, and tariff models are discussed, and become the basis for three axioms that underlie this framework. Driven by measurements and feedback over the longer term, the architecture of the network can also be tuned to meet the needs of the users, at least theoretically, with less reliance on over-provisioning for network services as a service strategy.

Once this framework is in place, tuning it periodically based upon usage experience and further research has the potential to make the overall pricing system both adaptive and capable of meeting its original objectives of network efficiency, technical feasibility, and user acceptance. There is a great deal of content in this paper, and it is worth a careful reading by those who seek to develop workable charging algorithms and fair pricing for services on the Internet.