If you can't bill for it, forget it ATM service providers face new billing challenges

As asynchronous transfer mode services quickly emerge to support growing volumes of users and traffic, it is clear that ATM technology poses unique new challenges for carriers and corporations. This is particularly true in light of the service provider's move to rapidly consolidate its networking infrastructure using ATM as the "great homogenizer."

Consequently, collecting information regarding network usage for billing and service differentiation is at the top of every carrier's hot list. But up to this point, it has not been clear how best to approach this issue. Also on everyone's mind is making the business case for an ATM network infrastructure.

To justify the cost of new equipment, companies must be able to leverage the infrastructure to provide new revenue streams and bring new products to market.

With essential standards in place and advanced equipment available for building production-quality networks, the industry is now taking a more pragmatic business approach to implementing ATM services. It is also quickly discovering how little it understands about the most significant aspect surrounding the implementation and operation of ATM networks: billing.

As campus deployment drives ATM into the wide area, carriers, switch vendors and users are being forced to consider the billing challenge. There are numerous recurring costs associated with wide area network services. Monthly usage charges can consume a significant portion of a user's network operating budget. Carriers providing ATM services with their own private wide area ATM backbones have even higher operating costs.

Before ATM services can take off in the WAN, a flexible, easy-to-understand and equitable billing process is required.

Usage-based billing has a strong value proposition for ATM. Previous tariffing in the telecom industry often has appeared to reward large users with high-volume discounts, while relying on low-volume users to subsidize the costs. These high fixed fees not only have priced small users out of the WAN telecom market, but also have acted as a barrier to market entry and the acceptance of new technologies.

Usage-based billing offers incentives to large users, while providing small users with the ability to scale their usage appropriately. Although usage-based billing clearly benefits carriers, enterprise providers and customers alike, its true value lies in its implementation. Today, there are three primary approaches to implementing usage-based billing: switched virtual circuits (SVCs), inter-destination usage or a hybrid solution of the two.

Switched virtual circuits Many view the virtual circuit, especially the SVC, as the foundation for usage-based billing. At first glance, the SVC looks promising because it is the basic means of using ATM. But the SVC has several drawbacks. One of these is that numerous SVCs and permanent virtual circuits are typically needed to support the connection that carries the application traffic. These are often referred to as infrastructure or maintenance SVCs.

Local area network emulation is one example of this. A LAN emulation client may need to make several calls to access the LAN emulation server before the SVC connection to other LAN emulation clients is established. This pattern also occurs as more sophisticated applications are deployed on ATM networks. Applications will access other resources, such as directory services, resource locators or policy managers, before the actual SVC carrying application traffic is established. Thus, users may be billed for multiple SVCs handled within the infrastructure of call setup. Without detailed information about the internal operation of each ATM application, customers may end up paying substantially more than they intended.

Another problem with charging on an SVC basis is that many of these infrastructure SVCs are of short duration. Ironically, collecting the information about the SVC and delivering it to a billing site may consume more network bandwidth and resources than the SVC traffic itself. This approach can lead to scaling problems as network usage grows, with the amount of bill- ing information outstripping the amount of data carried, especially for the infrastructure

SVCs. Multicasting applications pose yet another challenge for an SVC-based approach. Most multicasting services are dynamic; users or user groups "subscribe" for a multicast service for varying lengths of time. Maintaining quality of service with highly dynamic traffic patterns and demands places considerably more strain on the ATM network. Tracking and associating multiple end stations with a single transmitting source is also difficult within a single SVC perspective.

The inter-destination model A new approach deserving consideration is something called inter-destinational usage-based billing.

This approach measures the total traffic between two points over a specific period of time and uses it as the basis for billing. This usage-based approach is fair because users are charged for the amount of traffic they interchange.

Inter-destinational billing can be set for different time periods and different levels of service. Traffic in peak-time, heavy-usage periods, for example, can be charged at a higher rate than bulk transfers in off-peak hours. This strategy makes effective use of premium bandwidth during business hours and encourages customers to do intelligent application scheduling to take advantage of discounts for low-traffic periods. In addition, because ATM can support various levels of service quality, customers need to be tariffed appropriately by service level. Thus, high-performance SVCs providing minimal latency and high throughput can be charged at a higher rate than low quality of service for bulk transfer. In this way, inter-destination billing not only reflects actual usage but also takes into account traffic prioritization.

To implement this approach, the points used between inter-destination traffic must be well-defined. One metric might use ATM source/destination address pairs to collect traffic information for a particular service class flowing between two sites, while another might monitor traffic on a port-to-port basis from the input port on one ATM switch to the output port on the delivering switch.

Multicasting also fits with an inter-destination approach since the traffic between the source and multiple receivers can be tracked and measured over a period of time. Traffic volume can be adjusted so that users do not pay for multiple copies. The dynamic nature of multicasting is easier to handle since time is the only measure-users can subscribe and drop off as they need to without incurring additional SVC setup charges.

Finally, an inter-destination usage billing scheme must monitor the delivered traffic rather than the submitted traffic to benefit users. Any ATM network occasionally drops cells or loses SVCs with time outs; however, users should be charged only for the data delivered.

A carrier also may want to consider a hybrid approach that combines attributes from both inter-destination and SVC-based strategies. Some possibilities include identifying the infrastructure SVCs, placing usage thresholds on regular SVCs and providing criteria for switching between billing algorithms.

Infrastructure SVCs include calls necessary to reach directory services, routing services, multicast management services and other resources shared by a large user community. Some of these may be resources included within the ATM network itself or those provided by customers using it. Carriers may consider not tariffing infrastructure SVCs because they are usually of short duration and establish the virtual circuits supporting ongoing application activities.

Usage or volume thresholds could be applied to application SVCs so that low-volume customers are not severely penalized. For example, an SVC requires a basic setup charge; those resources are used regardless of the call duration or data volume transmitted. Below a certain threshold, the SVC could be billed only for setup, and the data volume would not be considered. This offers simplified billing and allows for cost containment for low-volume users. Another usage threshold could trigger a transition from billing on an SVC basis to usage billing on a flat-rate basis.

Any scheme that changes billing parameters must have links into the customer's own network management system. Private carriers must provide customer network management access to their own management with information so shifts in billing strategies are reported in real time. This integration and interaction ensures that customers can track their networking costs on an ongoing basis and have information to change their billing categories or select alternate vendors.

Now what? To make any of these approaches viable, carriers need a network architecture that provides ongoing capabilities for collecting information, monitoring traffic flows and adjusting billing algorithms appropriately. This architecture must be basic to the network while easily extendible through open integration interfaces. This will allow the extraction of standard billing information that can be processed and modified with carrier-specific billing schemes. In this way, the architecture of the network would not require modification as new billing requirements emerge.

Fundamental to the architecture is the ability of the underlying infrastructure and switching systems to collect, process and deliver huge volumes of statistics for each connection. Today, the statistics collection provided in most ATM switches is too rudimentary to support a usage-based billing system. More sophisticated collection techniques are being developed to provide higher granularity. For instance, traffic usage should be collected as delivered traffic statistics or on a per virtual circuit basis.

Tracking the infrastructure of low-usage SVCs requires even further granularity. Incorporating more intelligent agents into the switches allows local testing for threshold conditions with the appropriate signaling in the billing records. The signaling would indicate when thresholds have been reached, moving the customer from one billing strategy to another.

Also needed are efficient bulk transfer mechanisms for moving accounting information from switches to the accounting site to minimize the effect on the network. Other issues need to be addressed once the accounting information collected from the switches is delivered to a billing center.

While simple network management protocol can set the appropriate accounting, thresholding triggers and collection parameters, adding trivial file transfer protocol (TFTP) provides additional security mechanisms for protecting switch operations and accounting data from unauthorized access or modification. TFTP also provides a superior bulk transfer mechanism.

To harness the value of this information-all the billing and operational statistics, as well as customer account data-the architecture must be integrated. An integrated architecture allows information to be accurately analyzed, packaged and distributed to the appropriate parties. Some billing information will be accessible on-line through CNM interfaces, while other information will be provided on a regular basis to accounting departments.

Carriers and switch providers must work together to define a standard set of interchanged billing records and formats between switching products and accounting centers. This requires, for example, new automatic message accounting billing formats for exchanging information between elements of the network provider's system. As global ATM networks are deployed, these formats must be commonly approved to increase interoperability and exchange of billing information.

Getting the word out Usage-based billing provides the fairest, most equitable and easiest way for customers to use wide area ATM services, whether provided by carriers or by internal organizations. The specifics and particulars about different pricing schemes still must be determined.

Considering the appropriate architecture, and planning for a usage-based scheme, is necessary for both carriers and users in order to understand the implications of this next step in the evolution of the network.

Customers must understand the potential cost effect, not only for implementing an ATM network but for ongoing support and services. Lack of information increases the difficulty of accelerating ATM deployment. Without appropriate information, organizations will not have a solid business case and cost structures with which to evaluate different carriers or to estimate the operational costs of deploying an ATM backbone.

Peter Alexander is Executive Director of the wide area network business unit at Cisco Systems, Inc., San Jose.