Introducing Incentive Pricing Schemes in the ISP Market


Provision of Incentives Through Dynamic Pricing

The increasing demand for network service quality guarantees (e.g., low average delay rate, low packet loss rates) offers new business opportunities for the ISPs. This trend is intensified by the commercialization of new applications (e.g., IP telephony, videoconferencing) that have specific service quality requirements. Demand for network service quality may become the key driver to the adoption of incentive-compatible pricing mechanisms (Constantiou & Mylonopoulos, 2001). Assuming that most of the technical barriers for provisioning service quality will be overcome, the business relationships between involved market players will become critical. Pricing mechanisms based on incentives may induce an ISP to offer service quality guarantees. Prices should reflect the cost of service quality provision, as well as the opportunity cost of not serving or reducing the service quality to other users (Constantiou & Courcoubetis, 2001). Moreover, incentive-compatible mechanisms may induce efficient revenue distribution among ISPs.

Although incentive pricing approaches seem quite ahead of current market practice, there are strong market indications that in the near future such pricing schemes might be taken under serious consideration. In particular, many ISPs have recently introduced usage-based schemes to increase their profits. Congestion-aware pricing schemes are also being used through "time-of-day" pricing schemes (TheList, 2002).

Incentives provision can be achieved through dynamic pricing that enables the end user, who values highly a network service, to obtain it regardless of the level of congestion (Gibbens & Kelly, 1999). Price may change during the service delivery, based on specific criteria (e.g., congestion). For example, when a network exhibits high traffic, by increasing the price for the network service, customers can be obliged to reevaluate whether they want to continue the service or stop it. Customers with a lower valuation of the service will postpone their usage of the network. End users that choose dynamic pricing may benefit from a stable level of service quality independent of the time or the number of active end users on the network.

Dynamic pricing also enables ISPs to customize their services by offering them flexibility to create any pricing scheme according to specific customer needs. Besides, ISPs may charge their customers according to different payment models (e.g., receiver payment model, sender payment model, or a combination of both payment models).

Several proposals on dynamic pricing mechanisms' implementation have been made. MacKie-Mason and Varian (1994) proposed the "smart market" pricing scheme, which serves as a benchmark for other proposals. According to the scheme, at discrete time intervals end users submit bids for their packets to be sent. In the following interval the network carries those packets with the highest bids, limited by the network capacity. The price for the carried packets is set equal to the highest bid of the rejected packets. This scheme is incentive-compatible—users have no incentive to bid anything other than their true valuation.

Songhurst (1999) presented results from CA$hMAN project's trials of pricing schemes for the ATM available bit-rate (ABR) service. Network capacity is allocated on the basis of each user's declared "willingness to pay" (charge per unit time). As demand for resources increases, users must increase their willingness to pay to maintain the same utilization rate. In a related paper, Courcoubetis and Siris (1998) discussed the application of proportional fairness flow control to large ABR flow aggregates, where the appropriate notion of flow was the effective bandwidth of the streams. In this model, flows compete for "burstiness" transfer capabilities, and use the parameters of the ABR mechanisms to express their preferences for peak rate transmission, while the network uses the same mechanisms to convey price information back to the users.

Ramakrishman and Floyd (1999) proposed the implementation of Explicit Congestion Notification (ECN) in TCP. Resources that are nearing congestion can signal this fact to end systems by marking packets before it is necessary to start dropping packets. The authors of this proposal suggested that end systems should back off in response to marked packets. Gibbens and Kelly (1999) proposed that congestion marks should be interpreted as charges, and end systems should be free to vary their rate in accordance with their utility and the charges received. The theory identifies which packets should be marked so as to ensure incentive compatibility. However, it is not possible in practice to mark exactly the right packets since they precede congestion.

The main drawback of dynamic pricing is the variability of prices, which may annoy and discourage end users, who seem to prefer predictable prices (Altmann & Chu, 2001; Bouch & Sasse, 1999). From a technical perspective, incentive pricing schemes should avoid complicated mechanisms at the core of the network and place the "intelligence" at the edges instead, following the current Internet philosophy. In addition to this, there is an important tension behind the dynamic pricing approach—the debate between supporters of dynamic pricing and those that believe that over-provisioning of network capacity and simple pricing schemes (like flat-rate pricing) will be adequate to support acceptable service qualities.

Deploying Incentive Pricing Schemes

Due to fluctuating demand, resources are never plentiful everywhere in a network at all times. Over-provisioning can be an effective strategy, but mainly for backbone networks that heavily invest on infrastructure. Offering service quality involves decisions on how much network resource each user is entitled to. Traditional approaches have been focused on the management of network resources (Huston, 1999). Market-managed approaches argue that these decisions should be left to the customers—based on their needs and service prices.

A market-managed approach, where dynamic prices regulate bandwidth consumption in a single-class network, may be a better approach than simple over-provisioning. During the Market-Managed Multi-service Internet (M3I) project, new charging technology that enables the deployment of incentive pricing schemes like dynamic pricing has been developed. The charging technology also provides a set of generic pricing, accounting, and rating mechanisms at the network layer, which enables ISPs to charge for the usage of their network resources (Altmann, 2001).

There have been concerns on the cost and the complexity of implementing dynamic pricing, the stability and scalability issues of the resulting system, and whether the end users can effectively react to dynamic prices. During the M3I project, extensive simulations were made and demonstrated stability and scalability, while investigating the tradeoffs between stability and speed of convergence (Songhurst, 2002). The project provided a framework for tuning essential network parameters and for explicitly setting prices of congestion marks given the state of the network. The project developed simple algorithms for intelligent agents at the edges that address generic application requirements such as file transfers and multimedia. Simple arguments suggest that a self-managed network makes better use of its resources and reduces waste of bandwidth. ECN implementations are natural candidates for inexpensive pricing information propagation, making the above approach even more realistic (Johnstone, 2002).

M3I identified three new roles that support existing market players or enable the emergence of new ones. These players mediate in pricing interactions and are expected to alleviate complexities of dynamic pricing. The dynamic price handler emulates an end user's reactions to changes in connectivity prices, according to the strategy set by the end user. The risk broker enables connectivity providers, who use dynamic pricing, to offer simple pricing schemes for end-to-end network services. The dynamic price handler and the risk broker present to the end users predictable service contracts. The clearing-house assigns one single price for each IP session and distributes the revenue to all service providers involved in the service delivery. The price may cover either a single network service or a bundle, including an application service. The clearinghouse eases the management of contracts and the settlement of the payments for each service.

The Dynamic Price Handler

The dynamic price handler (DPH) automates the end user's reactions to dynamic prices. The DPH may be implemented as an intelligent agent at the end user's premises (e.g., residential gateway equipment). DPH makes choices on behalf of the end user, based on his strategy. DPH's objective is to solve a dynamic optimization problem for the end user by combining his preferences and his willingness to pay with the dynamically varying prices and/or network service quality.

The DPH interacts with the end user and the connectivity provider. The end user (typical receiver and paying party) communicates his strategy to the DPH. The DPH, according to this strategy, informs the connectivity provider, who is responsible for setting the prices according to traffic load and sends the data to the end user. Depending on the dynamically varying prices of the connectivity provider, the DPH dynamically varies network service classes according to the end user's strategy. The strategy is defined by the price and the service quality model (e.g., the transmission at a different priority level or peak bandwidth). At the end of the service's session, the DPH informs the end user about the accumulated cost to be paid.

The Risk Broker

The risk broker calculates prices for different network services by considering the fluctuations due to dynamic pricing. The connectivity provider offers services that may have highly varying prices or service qualities. The risk broker function takes part in the network service provision by mediating between the connectivity service provider and the end users. The risk broker offers a list of network services at certain prices to the end users at its own risk at the beginning of a service session. The network service/price chosen by the end user will be paid.

Assuming a network service market where connectivity service providers use dynamic pricing, a brokering service that offers service and price guarantees may be valuable to the end users. The risk broker role may be part of a market player that interacts with end users, and connectivity or application/content providers. The end user (paying party) communicates to the risk broker the choice of transport service and duration before the service session. The risk broker informs the connectivity provider on the end user's choice of service class. This information may also be communicated to the application service provider. Then, the connectivity service provider sets the dynamic prices and informs the risk broker. The risk broker offers the contracted transport service by using the underlying dynamically priced service of the connectivity provider. The risk broker also communicates to the end user a list of transport services and the corresponding calculated prices. The end user will choose one and pay the risk broker. Finally, the risk broker will pay the connectivity provider the charges accumulated through the network usage at different prices.

The end user needs such information in order to decide whether he should use a risk broker service or not. For example, an end user might have the choice of paying an aggregate cost for his usage once a month or a fixed amount that would be very close to the aggregate cost. Whether end user prefers to pay an aggregate price or to be "insured" depends on the distribution of the dynamic price and, in particular, on the probability attached to the event that the dynamic price becomes extremely high.

The risk broker function has some similarities to the dynamic price handler. The dynamic price handler allows end users to relate directly to dynamic pricing; the risk broker function takes part in the service provision by mediating between "traditional" services and dynamically priced network services. Risk brokering involves translation of usage incentives. The costs of risk brokering are fundamentally related to the differences in usage incentives offered and usage incentives faced.

The Clearinghouse

IP sessions may generate two or more bills. For example, network and information services may be billed separately. The task of creating an IP session with one price, which may be a bundle of network service with an application, or may be distribution of cost/ revenue in arbitrary manner among the end users, is the function of "clearing." In many cases two or more connectivity providers demand payment of a certain amount by the end users. The clearinghouse collects the demands from the parties to be paid, sums them, calculates the percentages for each paying party, and announces them the amount to be paid. These tasks are performed either after the network service has been delivered or online, during the service delivery. The clearinghouse expects all parties to pay the agreed share of the total cost of network service delivery without repudiation. Note that the total cost of network service delivery may not be known in advance.

The commercial exploitation of the clearinghouse role may increase as communicating parties start to share a volume-based cost of service delivery (Altmann, 2001). The clearinghouse is not responsible for the services provided to the paying parties. The natural function of the clearinghouse is the distribution of payment to the involved parties. The clearinghouse reduces the cost of multiple transactions, by aggregating money transfers to its regular customers and suppliers (connectivity providers and information providers). In addition, involved parties have to agree on a common mode of payment such as exchangeability of currencies or a form of micro-payments. A necessary condition is that the clearinghouse is trustworthy by the parties involved in the transaction. The means of establishing trust (e.g., Trusted Third Parties) are in the domain of e-commerce and outside the scope of this chapter.




Social and Economic Transformation in the Digital Era
Social and Economic Transformation in the Digital Era
ISBN: 1591402670
EAN: 2147483647
Year: 2003
Pages: 198

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net