Efficiency experts: By implementing ATM in both the switching and transport infrastructure, carriers can significantly improve network efficiency

Confronted by a rapid increase in demand for data networking services from business customers as well as Internet users, carriers are looking for a way to manage multiple services on a common infrastructure. In fact, integrating voice, data and video services over a single network has been a major carrier goal for many years.

Although data services clearly are driving industry growth, voice traffic still accounts for the majority of revenue-and bandwidth consumption. Currently, most carriers do not derive enough revenue from their data offerings to justify an investment in a stand-alone network. However, to handle the growing volume of data traffic, they rapidly are using up any surplus bandwidth they might have on the voice side of the network.

Although they cannot afford to maintain two separate infrastructures, prudent network planners know they must expand network capacity to accommodate new services. Any new equipment installations must support both voice and data applications. Although numerous technologies do so, only asynchronous transfer mode can be implemented in both the switching and transport infrastructures.

A modified model Looming on the horizon is the widespread deployment of asymmetrical digital subscriber line (ADSL) technology. ADSL will enable carriers to provide larger bandwidth pipes and move Internet calls with their long hold times off the local switch. Yet while it promises to solve one set of issues, ADSL also presents another significant challenge for carriers. Because many ADSL systems use ATM to transport data traffic, how can carriers manage ATM facilities within their existing asynchronous and Sonet infrastructures?

One aspect of this challenge is the need for broadband facilities. Most central offices today use DS-3 (44.74 Mb/s) or STS-1 (51.84 Mb/s) electrical connections as their largest broadband facility. While this is sufficient for DS-1 (1.54 Mb/s)-based services, it won't hold up under the demands of larger bandwidth services. Emerging high-capacity services are prompting carriers to rely more heavily on OC-3 (155 Mb/s) and OC-12 (622 Mb/s) optics for intraoffice broadband facilities.

Another challenge is broadband ATM facilities management. ATM, like Sonet facilities, requires certain network maintenance functions-grooming, private-line management and network testing capabilities.

Because of the magnitude and complexity of this challenge, carriers need a new network model, one that not only supports existing voice applications but also provides the infrastructure for growing data services. Currently, the network includes voice switches interconnected via a transport infrastructure. It is possible to design a similar and complementary topology by adding ATM as an extension of the existing network.

This unified network can support all voice, data and video services on a common transmission infrastructure, as well as interact with Internet protocol (IP) routers for Internet services. The new network model comprises three layers (Figure 1). Level 1 is the transport layer, Level 2 is the switching layer and Level 3 is the routing layer.

The foundation of the network is the transport infrastructure, which supplies the interconnectivity needed to support the switching and routing infrastructures. The transport layer is the most efficient, cost-effective means for interconnecting the various switching platforms that provide services to end users. In turn, this new network model is the most efficient, cost-effective means by which carriers can manage ATM and IP traffic.

The switch layer provides the call setup and tear-down functions throughout the network that are required to deliver services over voice, frame relay and ATM switching platforms. IP routing, for the delivery of data services, creates an additional layer above the network switching layer.

Using TCP/IP, carriers can provide Internet services over many different types of transport infrastructures-Ethernet, ISDN, frame relay, ATM or Sonet. For end-to-end customer service, IP relies on a connectionless protocol that requires the administration of routing tables throughout the network. This connectionless routing is most efficient when the data to be transmitted contains a relatively small number of packets for each end-to-end flow.

When carriers need to transmit a large number of packets for a single end-to-end flow, a connection-oriented protocol is more efficient. The carrier can set up an end-to-end path to transport all relevant packets without the processing required at each switching platform to determine each packet's destination from the router header.

The trade-off between connectionless and connection-oriented data transmission necessitates interaction between the IP layer and the switching layer, simply to ensure that network resources are efficiently used. Several methods for setting up the end-to-end connections for large IP traffic flows currently are under investigation. Ipsilon's IP switching and Cisco's tag switching are two examples that provide the interaction between the routing layer and the switching layer for the efficient support of IP traffic through large networks.

With today's network, voice and data customers are connected to the appropriate switching platform through the "network cloud," which provides a transparent interconnection for the switching platforms. The transport network performs its job so efficiently that most people dealing with the switching infrastructure hardly know it is there, yet it provides critical network functions.

The transport network efficiently and reliably interconnects the customer and the switching systems that provide customer-ordered service (Figure 2). Add/drop multiplexers (ADMs) and digital cross-connects (DCSs) further enhance the transport network's efficiency. ADMs interconnect the customer premises with the CO, as well as with all the COs in a carrier's network.

Ensuring the network's ability to survive for both linear and ring applications has long been a cornerstone ADM strength, enabling carriers to provide high-availability services to customers. Cross-connect systems primarily have served as vehicles for managing the interconnection of CO bandwidth.

Such bandwidth interconnection is necessary, for example, to support service churn and to consolidate the payloads of partially-filled asynchronous and Sonet facilities, thus more efficiently using switch port resources and interoffice transport bandwidth. The cross-connect has become the centerpiece of telecom networks, with nearly all the CO bandwidth traversing it.

The role of the cross-connect Because remote operations centers can prompt the cross-connect to redirect traffic, the cross-connect has become the primary vehicle for managing non-switched-or private-line-services and for maintaining the network. Economics have driven both cross-connect functions.

A lower price per port on the cross-connect caused private-line services to be migrated away from switching platforms, while carrier staff reductions created the need to provide remote test access functions on the cross-connect for turn-up testing and troubleshooting. The addition of ATM to the network has not eliminated the need for these critical transport network functions. Indeed, ATM actually has expanded the need to make the same functions available for asynchronous and Sonet equipment on broadband ATM facilities.

With the addition of ATM to the network, equipment in the transport layer must add ATM facility management functions to complement the management capabilities currently available for asynchronous and Sonet facilities. Specifically, ADMs must incorporate virtual path ring functionality to provide virtual path trunks and survivability for ATM traffic, while DCSs must incorporate virtual path and virtual circuit connection management to groom traffic for more efficient use of network resources.

Carriers can add ATM functionality one of two ways-by adding ATM functionality to the existing equipment or by collocating new ATM-based equipment with the existing equipment infrastructure. Each method has its pros and cons.

The major advantage of upgrading existing equipment is that it extends the economic life of the current investment. Such an approach also means the carrier does not have to create a new infrastructure to support new equipment; instead, the carrier can enhance existing methods and support systems to accommodate the new functionality of the embedded equipment.

Nevertheless, many carriers are choosing, at least in the near term, to deploy special-purpose ATM equipment, primarily because it is more readily available. Although this choice may be expedient in the short term, carriers will have to deal with several long-term issues, including scalability, reliability, maintenance and cost-effective management of private-line services.

Most special-purpose ATM equipment has been designed for the customer premises environment, rather than the carrier environment, and thus cannot scale to the hundreds-of-gigabits-per-second speed that cross-connects currently provide for asynchronous and Sonet facilities. In addition, special-purpose ATM equipment often does not satisfy carrier criteria for reliability and quality.

When carriers treat each service type independently in the transport layer, as they must with special-purpose equipment, they end up with networks that are very inefficient and difficult to manage-simply because each facility has unique properties. To minimize this, carriers today use cross-connects or ADMs to multiplex facilities from various switching platforms onto Sonet facilities.

In a best-case scenario, carriers consolidate separate DS-1 and DS-3 facilities onto a common higher-capacity Sonet facility. In the worst-case scenario, very little consolidation takes place, with each data network handled as dark fiber (Figure 3).

Both approaches significantly under-use network transport resources. Because peak loading for each type of traffic tends to occur at different times throughout the day, each dedicated facility has to be engineered for those separate peak times. On the other hand, a shared-transport medium means that a carrier has to support only aggregate traffic needs throughout the day and thus can size facilities more efficiently.

One of ATM's central advantages always has been its ability to support voice, data and video traffic-for switching as well as transport. This advantage first became evident in the campus environment, where most local area network and frame relay switching platforms have migrated to ATM at the switching layer.

The hybrid cross-connect In the carrier market, the network transport layer recently has begun to provide ATM functionality to support dynamic service requirements. In addition to providing multiplexing functions for ATM switches and ADSL, ADMs now pick up LAN traffic from routers and convert it to ATM for transport on an STS-1 or STS-3c facility through the access network.

The presence of ATM in the transport infrastructure also is driving cross-connect systems to support ATM traffic, again, because carriers want to integrate all traffic types on a common infrastructure. With the addition of an ATM switch fabric, a hybrid cross-connect system can groom ATM traffic from multiple sources onto a common backbone, thus boosting overall efficiency of the entire network (Figure 4).

Non-ATM services are converted to an ATM cell stream via the ATM adaptation layer (AAL) function, also referred to as segmentation and reassembly. Each service type has a distinct AAL function. AAL1 provides DS-1 adaptation, for example, while AAL5 provides LAN and frame relay adaptation. With the use of AAL functions, a single ATM backbone can efficiently carry all types of service, including voice.

The ATM cell streams are managed through the network within virtual paths and virtual circuits. The cross-connect provides the same functions for ATM facilities that it currently provides for asynchronous and Sonet facilities, including grooming, private-line management and network maintenance. Adaptation layer functions can be provided either within the cross-connect or at the edge of the network where the service originates.

With an ATM matrix, a cross-connect has the same capabilities as an ATM switch but goes two steps further: An ATM-equipped cross-connect provides low-cost management of private-line services and is optimized for scalability. Current Sonet cross-connects host up to 200 Gb/s of bandwidth.

An ATM-capable cross-connect can groom virtual paths for several partially filled facilities to create a new highly filled ATM facility. The virtual paths can be trunks between ATM switches performing real-time call processing or virtual paths handed off to the network for end-to-end private-line services. Switching ATM cells within the cross-connect matrix means greater efficiency for the entire network and lower operating costs for the carrier.

Within this architecture, the ATM edge switch-which may include an IP router-now receives well-packed traffic on its interfaces. Only those ATM cells that need ATM-switched service are routed to the edge switch, thus minimizing the size and maximizing the efficiency of that switch.

In addition, the only administrative tasks are those involving circuits that use the switched services. The hybrid cross-connect uses its ATM virtual path cross-connect capability to statistically provision a permanent virtual circuit to handle all other ATM traffic.

The cross-connect also provides a single point within the CO for monitoring and testing facilities. Carriers can monitor the performance of the Sonet and ATM layers of all CO facilities from one central point. For network maintenance, the cross-connect provides test access for the Sonet layer, as well as for the ATM layer, enabling carriers to troubleshoot the entire network.

By adding ATM capabilities to transport layer equipment, carriers will be able to migrate the network to a single ATM backbone that supports all services. This new infrastructure positions carriers to respond quickly and cost-effectively to the burgeoning customer demand for new and varied data networking services.