The virtual alternative

The Law of Consumption says that need expands in direct proportion to availability. Just as a raise seems to necessitate a higher lifestyle, more available bandwidth results in greater need for a high-speed telecom pipe.

Of course, price has a lot to do with perceived "availability." When video cameras cost $3000, people stuck with their Instamatics. Now that they sell for as little as $300, few households are without one. Following the same pattern, most businesses today access the Internet via 56K frame relay lines, and most individuals access it the cheapest possible way: using a dial-up modem over a POTS line. But as the cost of high-speed, broadband access becomes truly affordable, it will become increasingly essential across the board.

But that won't happen as long as Sonet is used as the network transport technology. Even though digital subscriber line (DSL) and Sonet technologies make access faster than ever, there's no disputing that Sonet was designed to carry voice traffic, not a mixture of voice, data and video. By constantly increasing Sonet transport capacity, service providers can meet the demand for broadband service in the short term. But the sheer cost of installing and maintaining this type of network will limit consumer demand for broadband access.

The solution is to use asynchronous transfer mode technology - not just for switching but for transport. After all, ATM was designed from the ground up for high-speed multiservice communication. A new ATM transport element makes it possible to do just that. This new network element combines the transmission speed and reliability of Sonet with the primary features of ATM, such as the ability to manage bandwidth efficiently and integrate multiple services on one infrastructure.

Using bandwidth more efficiently

Broadband services provided over Sonet-based networks are expensive because Sonet technology uses bandwidth inefficiently. Time division multiplexing (TDM), the standard transport technology, requires carriers to provide each customer with a discrete amount of dedicated bandwidth - enough to meet their maximum usage requirement.

As a result, customers who want broadband service often must pay for a dedicated line from their facilities to the ATM switch - even though most customers' average bandwidth usage is only 10% to 15% of their maximum usage.

In contrast, a network that uses ATM transport elements allows customers to share bandwidth. Through statistical multiplexing, the ATM element allocates to each customer connection exactly the amount of bandwidth required at any given moment. As a result, carriers can serve more customers using the same amount of bandwidth. More specifically, they can oversubscribe bandwidth in inverse proportion to average use.

For instance, if traffic engineering studies show that users' average peak period bandwidth usage is only 10% of their maximum usage, the carrier can assign 10 customers to that amount of bandwidth. When one customer is sending an e-mail, which uses little bandwidth, another customer may be downloading a video clip, which uses much more. In the end, the majority of the bandwidth is occupied most of the time by one user or another.

Because the ATM transport element can perform statistical multiplexing, customers no longer have to pay for a dedicated line from the end user to the broadband switch. Carriers can sign up more customers using less bandwidth and fewer switches, and more consumers discover that they can't live without broadband access.

Share and share alike

The ATM transport element also reduces carriers' costs by allowing them to transport multiple services on a single network infrastructure by using integrated service access multiplexing, as defined by Telcordia Technologies Inc.'s (formerly Bellcore) TR-2842 standard. These costs fall into two categories:

Equipment costs. Traditionally, each type of broadband service - such as frame relay, ATM/cell relay, DSL or transparent LAN service - has required its own layer of switches, network management and backhaul facilities. In contrast, the ATM transport element converts all traffic types to ATM cells, which means that only one network is necessary no matter how many broadband services are offered (Figure 1). Obviously, carriers save a lot by not having to buy additional equipment every time they provide a new service. In addition, it enables them to offer new services at a faster rate and attract more customers.

Service provisioning. The ATM element's multiservice capability also saves carriers money in service provisioning. That's because predicting customer demand for a telecom service has always been a risky proposition. If a carrier undercalculates, customers become unhappy because their needs are not adequately met. If the carrier overcalculates, the organization may install costly equipment that becomes obsolete before it is fully used.

ATM transport element technology makes capacity planning far less risky. A carrier that miscalculates demand for a certain service can simply change an access interface card to shift more network capacity to another service that is in greater demand. The resulting cost savings apply to the customer as well. Rather than ordering one dedicated line for frame relay and another for private line, the customer can access both services over a single virtual connection.

The QOS factor

When carriers can better predict and accurately prepare for demand, they can also offer competitive service level agreements (SLAs). The ATM transport element helps carriers handle different types of traffic in accordance with ATM's quality of service (QOS) standards. While Internet protocol (IP) currently treats every transmission the same whether it is an e-mail message or a two-way videoconference, ATM can assign different priorities to different types of traffic.

For example, full-motion video or voice transmissions, which don't look or sound good if packets are delayed even slightly, receive constant bit rate priority. Data transmissions, which are far less sensitive to small delays, are given variable bit rate priority. And non-critical transmissions are assigned the cheapest virtual path and given an unspecified bit rate priority. As with IP, the latter type of transmission is simply sent again if the communication doesn't go through.

The ability to prioritize traffic enables carriers to offer customers end-to-end SLAs based on their transmission requirements. This is a significant competitive advantage.

Network reliability is another important aspect of QOS. When used in a virtual path ring architecture (as defined by Telcordia's GR-2837 standard), the ATM transport element provides the same superior reliability as Sonet. For example, if a facility failure occurs on any part of the virtual path ring, traffic is automatically switched to the counter-rotating fiber in less than 50 msec. If a failure occurs at the level of the ATM element, affected virtual paths are switched to protected units. Even placing ATM switches at the same locations as the ATM elements would not provide this level of protection from downtime and delays.

ATM transport elements have great architectural flexibility, fitting just as well in a small point-to-point network (Figure 2) as in a large ring configuration. This is good news for competitive local exchange carriers, which need to start small but must be able to expand rapidly when growth opportunities arise.

For example, by using ATM transport element technology, a small CLEC can lease only as much bandwidth as it needs on an incumbent LEC's network. The CLEC can then provide services to different locations with different bandwidth requirements. As it expands, the CLEC can graduate to an inexpensive DS-3 line terminating at its own central office. Then, the growing carrier can change a card to triple its capacity with an OC-3 line or fiber. Eventually, by procuring fiber, the CLEC can upgrade to OC-12c and OC-48c virtual path rings as demand requires.

Through all these transitions, the ATM element platform remains the same; only the interface cards change. ATM transport element technology may well be CLECs' best friend, allowing them to minimize equipment outlays and to grow incrementally as their customer base expands.

The obvious alternative

Today carriers see only two network architecture alternatives as they prepare to deliver more broadband services to their customers, neither of which is ideal. The first is simply to increase the bandwidth on their backbone transport networks. This involves adding more point-to-point lines to provide frame relay, cell relay and private-line service, or to connect to DSL access multiplexers (DSLAMs) at or near customer sites. Using TDM, each DSLAM concentrates traffic onto a high-capacity line that requires a dedicated port on the switch. Needless to say, the cost of additional lines, DSLAMs and switch ports adds up quickly.

The second alternative is to install ATM switches closer to customer sites. The switch itself then aggregates traffic before it hits the backhaul fiber, which allows the fiber to be used more efficiently. The problem here is the cost of full-service ATM switches, which are expensive to say the least. In addition, trying to connect ATM switches into a "pseudo-ring" comes with a whole new set of administrative, provisioning, re-routing, latency and timing problems. Using switches purely as transport shepherds is a tremendous waste of equipment resources, not to mention money.

In fact, money is the big issue here. If the only objective is to get broadband traffic from here to there, the old Sonet system works fine. It's fast and reliable. But if the goal is to make broadband services available at a price consumers can afford and carriers can reasonably absorb, the Sonet architecture is irretrievably flawed.

The solution, of course, is the ATM transport element. A typical metropolitan carrier shows why.

This particular model has 10 COs, one frame relay switch, and 1000 broadband circuits (Figure 3). The COs are connected on a Sonet ring via lines of different capacities (including DS-1s and DS-3s). Each line requires its own dedicated virtual path around the ring, as well as its own dedicated port on the frame relay switch. To serve the carrier's current customer base, the network requires 1.37 Gb/s of bandwidth and 330 ports on the frame relay switch.

Compare that with Figure 4, which shows how the same carrier could use ATM transport element technology - such as ADC Telecommunications' Cellworx service transport node - to meet the same requirements. This time the 10 COs are connected on an ATM virtual path ring using the service transport nodes.

Rather than following rigid TDM-based dedicated paths, customer traffic shares available bandwidth. Traffic priority is determined using ATM QOS levels. Because the ATM service transport nodes can adapt all types of traffic to ATM cells, the carrier can offer other kinds of broadband services as well, including cell relay and asymmetrical DSL. Network and service interworking are handled by a single ATM switch.

The bandwidth requirement for the latter type of network is 205 Mb/s, terminated on just two ATM switch ports. The total capital cost savings for a network using the service transport nodes is 55% lower than the traditional model, an estimate that includes transport and switching equipment costs.

The cost savings will grow even more as the carrier's network expands. Even if network service requirements double, the network based on service transport nodes will still be able to transport all the traffic on the same ATM virtual path ring. In contrast, the TDM method would require a second Sonet ring, as well as twice as many ports on the frame relay switch.

Clearly, the ATM transport element is a technology whose time has come. While maintaining the superb reliability and speed of Sonet, it significantly reduces the cost of broadband access. Carriers that pass these savings on to consumers will find hordes of new customers knocking at their doors, eager to obtain the new lifestyle standard: high-speed broadband services.