Charging for ATM-based IP Multicast Services

Charging and accounting of Internet services has been addressed in numerous proposals. However, finding a solution suitable for ATM-based IP multicast services is highly challenging. IP multicast over ATM introduces high complexity, in particular for multicast with heterogeneous QoS and separate administrative IP domains by involvement of multiple Internet service providers. This paper proposes a framework for charging of IP multicast over ATM that is suitable for heterogeneous scenarios with multiple interconnected IP service providers and non-negligible losses. A scheme for ATM-based IP multicast transmission of wavelet-encoded video is presented that uses multiple ATM service classes. The scheme supports multicast scenarios with heterogeneous QoS and allows to trade off video quality, error tolerance, end-to-end delay and costs. 1. Charging and Accounting Charging and accounting of IP services gain more and more interest as the Internet is moving from an academic to a commercially dominated environment. A satisfying charging model has to deal with the following three main aspects of charging: 1. Customer Aspects: The customer has to understand the charging model and must be aware of the costs he produces when using a service. The costs should not differ significantly from costs of the telephone tariffing system the customer is used to. This introduces a threshold of acceptance. Costs should be based on goodput, i.e. errors within the network that require additional traffic for error recovery should not lead to increased costs for the customer. 2. Economic Aspects: The model has to differentiate between several types of service and – within these types – between different levels of service quality. Thus, the customer is animated to choose the cheapest type of service to satisfy his needs. The costs can be assigned to the customer. 3. Technical Aspects: In order to minimize the effort for computing the different tariffs the charging scheme has to be as simple as possible. After having distinguished between different types of services the charging model can be optimized for each type. As cell-based charging in ATM networks is computationally demanding, dedicated hardware-support may be required. To summarize these three aspects, charging of ATM-based IP multicast services should be kept sufficiently simple to be feasible and intelligible, while allowing to differentiate between different types of service.