Implications of Introducing New Charging Technology

In this section, we briefly describe two generic scenarios about the commercial exploitation of the new roles. In the first scenario the dynamic price handler (DPH) describes the case of combining various service quality levels with simple prices at millisecond time scale and for different services using the customer's software agent. Then in the second scenario, the guaranteed stream provider (GSP) offers a retail network service with quality guarantees, which is synthesized from a wholesale service identical to that in the DPH scenario. Finally, we discuss the implications of introducing dynamic pricing in the ISP market.

The scenarios are based on a generic framework that enables the creation of innovative pricing schemes. We assume the existence of a QoS mechanism (e.g., RSVP signaling or the ECN marks). ECN marks are interpreted as price and/or load signals that generate a packet-marking rate. A gateway analyzes this rate and determines what price to charge for ongoing and new services. The fundamental factors of this framework are:

• The IP network middleware on customer and provider systems that gives real-time control over application and network quality during business transactions (e.g., buying, selling)

• The middleware between providers and customers, along the value chain, that enables them to switch between different QoS technologies and pricing schemes

• A management approach for network resources, based on pricing, even if it is hidden from the customers

Dynamic Price Handler (DPH) Scenario

This scenario explores the concept of a dynamic price handler agent on the end user's equipment reacting to dynamic pricing (ECN marks). ISPs have to deploy dynamic pricing software on their networks. In particular, they may deploy ECN on all routers and set the congestion experienced field in the IP packet header with a probability related to current load on the egress interface as well as a charge per mark (Johnstone, 2002). The motivation is to give the end user a price incentive to react to congestion, while allowing him to pay to ignore a certain level of the incipient congestion, if the value gained is greater than the charge levied.

The receiver's ISP then offers network service at a price calculated by placing an effectively fixed price on each ECN mark (the same pricing scheme can also be used between ISPs). To avoid increasing pricing for low-quality service, the marking rate should rise just before congestion appears (Johnstone, 2002). ISPs do not insulate their end users from variable qualities or prices. Instead, end users insulate themselves from this unpredictability with the DPH agent. The DPH agent optimizes the end user's utility from service within the constraints of his strategy per task, which are either required from a specific application or set by the end user at the beginning of a service session. Each strategy is a small data object that encapsulates how utility varies with bit rate for each task (Andreassen, 2002).

Figure 4: Dynamic Price Handler (DPH) Scenario

Different end user strategies can give each agent complete control over the network behavior of the various sending applications. Responses to congestion range from completely elastic (like TCP) to a completely inelastic "non-response," holding a constant bit rate by paying whatever is necessary during congestion episodes up to a threshold (self-admission control). Strategies between these two extremes provide the flexibility to move the bit rate to whatever is considered best value for the task in hand, given prevailing congestion conditions. Agents controlling flows through the same bottleneck interact, intermediated by congestion signaling. While some inelastic agents are paying to hold their rates, the more elastic agents back off in order to avoid paying.

DPH offers price predictability to the end user while demanding his higher involvement. DPH gives more experienced users the ability to achieve their own service quality and cost objectives without sacrificing the simplicity of the network architecture.

The Guaranteed Stream Provider (GSP) Scenario

Applications such as IP telephony as well as real-time audio and video have high service quality requirements. The provision of communication, classical telephony-like service to end users through IP networks, is the motivation for this scenario. In this case the total ISP's benefit is greater if some users are blocked out while the rest of them are given expected bandwidth capacities to serve their applications. Unlike agent-based self-admission control in the abovementioned scenario, the GSP described in this scenario offers admission control as an extra service. Pricing of capacity reservations can be completely static, or the ISP may also choose to vary reservation pricing on slow time scales such as time of day (Andreassen, 2002).

ECN marking is deployed on every router, as already described in the DPH scenario. The GSP uses RSVP-enabled routers only at the edges of the network, and reservations are forwarded between them across the ISP's non-RSVP-enabled routers, which simply treat them as data. The signaling interface between ISP and GSP is simply the packet-marking rate (Johnstone, 2002).

Figure 5: Guaranteed Stream Provider (GSP) Scenario

According to the GSP scenario, two market players cooperate in order to provide the service to end users. These market players are: the ISP, which provides a basic network service, and the guaranteed stream provider (GSP), which has the means to offer guaranteed services. The GSP incorporates both the risk broker and clearinghouse roles. The ISP deploys dynamic pricing on the entire network. The GSP protects reserved traffic across the ISP's network by simply clearing the ECN capability field in the packets of unreserved traffic. During congestion, unreserved traffic is dropped while reserved traffic in the same link is congestion marked (Andreassen, 2002). When congestion signaling rises above a threshold, the GSP starts denying admission to new reservation requests. Thus, the level of reserved traffic across the ISP is kept below capacity because signaling starts before traffic load reaches congestion.

The observation that end users prefer simple contracts with a predictable final charge is a crucial parameter for the commercialization of the GSP. The GSP mediates between the proposed efficient dynamic pricing mechanisms at the network level, and the simple and predictable service interfaces that end users appreciate. Service differentiation within a network may have unpredictable cost, especially in case of congestion. The GSP offers to end users simple and guaranteed service quality contracts, alleviating the complexity and instability of the respective service quality mechanisms of the network.

Implications for the Market Players

According to the ISP reference model (Figure 1), new pricing and charging technology may affect several business relationships. An information service provider that delivers content services based on multimedia applications (e.g., real-time video) has specific service quality requirements that cannot be satisfied without fair compensation of the cost incurred by backbone providers. Thus, dynamic pricing may increase the cost of network services for the information service provider, but at the same time will generate higher revenues from the customers that will receive the content service.

In addition to this, information service providers may benefit from dynamic pricing of differentiated infrastructure services offered by data center providers by bundling of different network services with information services. The data center provider may reduce its costs by allocating resources according to the information service providers' needs and willingness to pay. Dynamic pricing also enables the data center provider to charge information service providers according to the consumption of resources, as well as the congestion that they are causing.

Furthermore, access and backbone providers can use new charging technology to differentiate their network services and to deal more efficiently with network congestion.

End users adopting dynamic pricing will benefit from constant quality of service— independent of the time of day—or traffic load on the network.

The deployment of new charging technology, which enables the introduction of dynamic pricing schemes, also allows the provision of new services. As an example, we present possible changes in the business model of AOL-Time Warner and Exodus.

AOL-Time Warner purchases network capacity from backbone and access providers, and sells it to its end users. In case backbone providers offer dynamic pricing schemes, AOL, by incorporating the risk broker role, can absorb dynamic price variations and offer to its end users stable prices based on their preferences for service quality and willingness to pay.

In addition, AOL-Time Warner—which has a large customer base and uses flat-rate or tiered usage-based pricing schemes—may expand the offered pricing schemes by providing premium services, which are charged on a usage basis. In case of network congestion (the demand is larger than AOL-Time Warner's purchased capacity), AOL-Time Warner will dynamically increase prices for the usage of network service in order to reduce the demand. End users under flat-rate pricing plans will be affected by a reduction in network capacity and, therefore, in service quality. However, end users under dynamic pricing schemes will be able to keep reserved capacity at higher, though predictable prices, through the use of the dynamic price handler. AOL will generate revenue from end users that are willing to keep a high level of service quality. When there is no congestion, AOL-Time Warner will generate revenue from end users that chose flat-rate pricing schemes. This flexibility in choosing pricing schemes might be very attractive to end users, since they can meet their needs with customized pricing schemes and without unpredictable costs.

Exodus may adopt the role of a clearinghouse. The clearinghouse role may enable Exodus to provide end-to-end service quality, even on network resources that it does not own. Exodus can share revenue from the end user or the information provider with all connectivity service providers involved in an end-to-end service, depending on the quality delivered. Revenue sharing may give incentives to connectivity providers to join in and provide high service quality.