For the past 50 years, the network circuit has been the building block of the telecom industry. Carriers own the copper wire and optical fiber, and they have the right to use the electromagnetic frequencies employed in wireless communications. With this media, carriers create digital circuits that transmit bits and analog network circuits that transmit voice and fax calls from one location to another.The price of all transmissions is based on metrics such as time of day, length of call, distance, bandwidth, latency and bandwidth asymmetry. (Asymmetric bandwidth, where the volume of data going downstream to the user is greater than the volume going upstream, is less valuable than symmetric bandwidth, where the same volume of data moves downstream and upstream.)
Although analog circuit-switching assets depreciate very slowly and ongoing maintenance costs are relatively low, recently per-minute fees throughout the world have been high. This situation has created today's highly profitable telecom market. However, due to changes brought about by global deregulation of telecommunications and advances in technology, that picture is changing. Deregulation has spurred the formation of competitive local exchange carriers (CLECs), while the widespread use of personal computers (PCs) has spurred the exploding use of corporate networks and of the Internet protocol (IP)-based Internet. Thus, IP-based networks are one of the major reasons for the growing demand for telco services.
The result: Fierce competition is making bandwidth a commodity. In most states it now is possible to buy an asymmetric line with close to 1 megabit per second (mbps) performance for only $200 a month. In the Washington area, a 45mbps circuit can be leased for $5,000 a month. International telephone calls are priced as low as 15 cents a minute. These prices are only about a tenth of what they were just a few years ago.
The Current Internet Economy
As the price of a circuit has gone down, so has the customer's perception of its value.
Most network users today are unaware of the bit-transfer circuit technology that underlies their communications; they simply perceive the higher-level IP services such as e-mail, database and e-commerce transactions, web pages, file and fax transfers, streaming video and voice over IP (VoIP). Surprisingly enough, customers of Internet services are charged one low, flat fee, based on the capacity of the circuit that connects them to the Internet service provider (ISP) and that encompasses all services. Second- and third-tier ISPs that connect to upstream ISPs also pay a flat fee for services. Flat fees are the norm because there currently is no industry-wide consensus on performance metrics that would allow customers to evaluate the price and performance for the level of service that they get when using their favorite IP applications.
Until recently, this inability to measure network performance has not posed much of a problem. Even when network performance is asymmetric and unstable, varying with factors such as time of day, network congestion and other unpredictable factors, customers still can have a satisfying experience with services such as web surfing and RealAudio most of the time, in most network conditions.
New Services Entail Changes
Interactive voice and video communications, on the other hand, are symmetric streaming services that use substantially more resources than other applications and require stable, predictable, symmetric performance. If VoIP, in particular, is to be deployed over the Internet, the Internet infrastructure will need to change. The largest market for VoIP is in the corporate arena, which can achieve considerable savings compared to analog voice. VoIP also is a cheaper, middle-tier alternative to long distance signaling system 7 (SS7)-based switched telephony technology. Most importantly, VoIP provides a way to avoid settlement costs related to the initiation and termination of international and long distance calls.
Over a corporate network, VoIP performance is reliable because the network is well controlled and bandwidth can be over-capacitated--as long as the wide area network (WAN) does not contain too many expensive long distance bottleneck segments. VoIP over the congested public Internet, however, can be an unpleasant experience with frequent interruptions.
Currently, the economy of the Internet is based on the fact that traffic is asymmetric and that packets are transmitted in a "democratic" environment; no packet is discriminated against or given a higher priority because it has been given a certain economic value. The asymmetric applications that run over the Internet today do not require the high level of quality of service (QoS) needed by applications such as VoIP that demand stable symmetric streaming. The networks of most ISPs do not have the capacity to deploy VoIP widely--which, as a symmetric application, consumes much more bandwidth than asymmetric applications--and still maintain a fully democratic packet policy. Nor at the typical $20-per-month flat fee can the ISP be compensated for the use of long distance bandwidth and the upstream peering fees incurred by VoIP.
Thus, ISPs are faced with a dilemma. Is there a way to deploy VoIP and other streaming symmetric applications over the public Internet--and still make a profit?
Yes, if service providers deploy usage-based IP metering and accounting systems to create an end-to-end usage-/cost-based Internet economy.
Cost-Accounting Challenges
The switched-circuit and telephone industry long has used what are called customer care and billing (CCB) systems to track telephone use and help generate those familiar telephone bills that itemize the destination, duration, class of service (CoS) and price of all calls. It makes sense to bring the potential of these CCB systems, for which there is a $3 billion market today, to the Internet. By feeding CCB systems with IP usage data similar to the call detail records (CDRs) they receive from circuit switches, service providers can create bills for IP services. A new class of solutions called IP mediation systems is emerging to handle that function. These systems collect usage information from devices throughout the network, format it and feed it to the CCB in the form of service detail records (SDRs) that are the IP equivalent of CDRs. In turn, the CCB is able to activate and provision user services via the IP mediation system.
This concept may sound simple, but there are many challenges involved in collecting usage data in a typical large IP network, and the IP mediation system must be able to deal with all of them.
Lack of Standards
One major barrier to usage-based billing has been the lack of standards. In a typical IP network there are many elements that terminate or mediate traffic, including routers, switching hubs, firewalls, web servers, proxies, multicast servers, name and directory servers and more. Each element is configured differently and has some sort of log or accounting mechanism--and each provides accounting data in different formats and provides different types of session accounting records. Sometimes the logs of such network elements are not readily available, making it necessary to use complex heuristics to generate the IP SDRs. A web server, for example, produces access log files that account for each access to the server. Some switches use remote monitoring (RMON), which provides state information and statistics about network traffic and which is accessible via simple network management protocol (SNMP). It is possible to transform such state information into transaction information.
There have been and are some attempts to create standards for service provisioning, session-based accounting and network metering. Remote authentication dial-in user service (RADIUS), developed by the Internet Engineering Task Force (IETF), is the current standard protocol for user authentication, and it also includes event accounting. A new IETF standard proposal called "Diameter" will enhance RADIUS and enable more general accounting abilities. Other IETF working groups in this field are the Authentication Authorization and Accounting (AAA) group, which is attempting to provide a generalized solution to the problem, and the policy group, which deals with a policy language.
The issue of network policy-based behavior, especially security and QoS, is of particular interest to many vendors in the IP market. Some vendors actually allow their device or server to be configured by an externally accessible database that uses the lightweight directory access protocol (LDAP) to communicate policy information. The Desktop Management Task Force (DMTF) has developed the directory-enabled network (DEN) specification, which allows network elements to receive policy information from a central directory server (presumably using the LDAP or common open policy system [COPS] protocol). Other bodies such as the International Telecommunications Union (ITU) and the European Telecommunications Standards Institute (ETSI) may attempt to step into this arena at some point.
Accounting records cannot have a fixed format in the IP environment because of the diversity of needs. In the telco world, each new service has to have some accounting mechanism as a prerequisite. In the IP world, accounting has never been a prerequisite and has always been associated with debugging and alert notification. Any accounting standard that evolves in the IP world in the near future probably will be content-neutral and focused on the transport mechanism.
Currently, however, no one standard has emerged that makes available all of the pertinent information an IP mediation system requires from all network elements. To create the IP SDR, the IP mediation system has to have some way to deal with this lack of standards when it contacts each network element to extract its transaction records. It is thus imperative that such a system be extraordinarily flexible, able to collect session information in a variety of formats, yet still ensure that each SDR provides consistent information to the CCB. And, of course, the system has to be efficient, scaleable to any size network and accurate.
Who's Who?
User identification is another important billing issue, because if you can't identify users of network resources, you can't bill them. In the telco world, phone numbers immediately are associated to billable entities. In the IP network, however, addresses may be assigned to customers dynamically, and there may not be a permanent user identification in the accounting data collected from the network elements. For example, when a dial-up customer logs in to the ISP's network, he or she calls a lead telephone number, but the call is terminated by the next unoccupied modem line in the ISP's modem pool. The IP address assigned to the customer usually is associated with that modem or chosen out of an address pool. Subsequently, it appears to any Internet server and router that any customer activity originates at that dynamically chosen IP address, making it hard to associate the address with the user.
The dynamic nature of information in IP networks does not end with user identification. For example, the domain-name service (DNS) is an Internet distributed database that associates IP addresses to host and domain names, and vice versa--and the association changes daily. This loose association of name and address also is a problem with dynamic host configuration protocol (DHCP), which allows a computer to be plugged into an IP network and get an unused IP address. This makes it difficult to identify a computer or its owner based on the IP address alone. An IP mediation system thus must associate data with a user in real time, before data changes.
Handling High Volume
The sheer volume of transaction information generated in an IP network far surpasses the volume generated in a telephony network and presents another set of problems. The myriad devices in the network of a typical network services provider (NSP) generate millions of transaction records per hour. Sending all of these accounting records through the network to a single database is not cost-effective and actually can jeopardize network performance.
To prevent this data from overwhelming the network, the IP mediation system needs to have a distributed, real-time, policy-based data filtering and aggregation scheme that extracts only user-specified billable records for further processing. The system then has to enhance this IP session data with information collected from other sources--such as RADIUS, DHCP, LDAP and domain-name servers, to generate complete, meaningful billing records.
Another problem arises because different network devices can record the same event. For example, a web proxy/cache and a router may account for the same web transaction. Since this data is used to bill customers, an IP mediation system must be able to recognize duplications and eliminate them.
An IP mediation system also must be able to deal with network transactions, such as a VoIP session, that extend beyond the NSP's network to run across several ISPs. Such transactions present a particular problem for ISPs that commit to a certain QoS, since they can give the commitment only within network boundaries. If a transaction takes place between a customer on Network A and a remote entity on Network B, the NSP cannot make a commitment for end-to-end quality, throughput and latency. Thus, an IP mediation system should be able to account for metered QoS performance at a transaction level and allow for discounting if the QoS is not provided.
An IP mediation system is an important element in making the convergence of voice and data networks work. The Internet and wide-area IP networks, which both carry the valuable application layer and content, are built on top of the switched circuits that carry the transport layer. Since the value the customer perceives resides at the application layer, and since carriers' transport-layer revenues are not as safe as they once were, it makes sense for carriers--the circuit owners--to want to own a piece of the action at the application layer by becoming ISPs or IP carriers.
It is not easy for a carrier to become a top-tier ISP. Providing good Internet service is technically challenging. Connecting with Internet backbone providers, network access points (NAPs) and metropolitan area Ethernets (MAEs) is expensive. Carriers have a large investment in CCB and operating system software that are unfit for IP networks and services. The ISP flat-fee billing model is not as profitable as leasing circuits--especially when it comes to offering VoIP in an environment that does not have policy-based QoS.
Thus, it makes sense for a carrier that wants to enter the ISP market to merge with an existing ISP that has technical Internet know-how, rather than build a network itself, and provide the ISP with the bandwidth it needs. The carrier incurs a one-time circuit setup fee and has few recurring costs. This provides a major advantage over independent ISPs, which have huge recurring bandwidth fees.
The carrier also has another advantage over independent ISPs because it already has an expensive CCB system that ISPs could never afford. The carrier can use the CCB, together with an IP mediation system for collecting usage information, to bill for expensive VoIP traffic on a per-usage basis. By deploying traffic shapers or other IP QoS technologies, the carrier can put a higher priority on the packets of higher-paying VoIP customers than on less-profitable packets. A merged carrier/ISP thus can offer not just connectivity but both inexpensive unmetered services and, via its IP mediation system, profitable, prioritized, high-end metered IP services.
Enabling Service Creation
As more sophisticated applications become available over the Internet, it is becoming clear that one-low-price-fits-all no longer is an appropriate billing model. Yet given the Internet's complexity, until now it has been impossible to create the IP equivalent of the telephone bill. This unsatisfactory situation is poised to change with the recent emergence of IP mediation systems, which will enable ISPs to charge customers accurately for their Internet usage. These systems also make the Internet market attractive to carriers, since the data collected by the systems and fed to their existing CCB systems can be used to bill for Internet services profitably.
In fact, the usefulness of IP mediation systems promises to go well beyond enabling user organizations to increase profits by billing for actual use of the IP network and services. With the IP session data collected by such a system, ISPs can create innovative, flexible pricing models. And they can use these models to develop and offer a broad range of profitable new network services quickly, which would be impossible to offer for a flat monthly fee. So don't be surprised if sometime soon ISPs start offering their customers an Internet videoconferencing "friends and family" package.
Limor Schweitzer is chief technology officer at XACCT Technologies Inc., Santa Clara, Calif.
