BY TONY RYBCZYNSKI

Why continue to invest in dead-end solutions? Enterprise users with T1 mux networks are faced with this question.

Enterprise users need to optimize the performance of their enterprise networks, while simplifying their operations environments, managing their capital investments, and minimizing their recurring costs. Consolidation of multiple networks is one element of the strategy to meet these challenges.

Network consolidation can deliver the highest performance for the dollar and simplify the networking infrastructure. Traditionally, network consolidation was achieved through T1 multiplexers, which have been widely deployed in enterprise networks. However, T1 multiplexers are optimized for a 64-kbit/s private line world.

T1 mux networks served users well during the 70s and 80s -- when the world was dominated by narrowband terminal to host point-to-point and point-to-multipoint circuits, and when anything above T1 was prohibitively expensive. But with two-digit per annum traffic growth from bursty LAN-based applications, and with the improved price-performance of public network frame relay and ATM services at multimegabit/s rates, T1 multiplexer networks no longer seem so fit. If they are not DOA already, they are -- at best -- in the ICU, given that investment is being capped and that parallel networks are being established. So what's next? Which network consolidation architecture fits today's and tomorrow's enterprise networking needs in a multi-Mbit/s LAN and ATM world?

REQUIREMENTS FOR WAN NETWORK CONSOLIDATION
Let's step back and take a look at user requirements. At a high level, the requirements are for responsiveness to traffic growth and new application needs, reliability (as business and network operation become inseparable), security, and cost control. Put another (and even more general) way, users need to maximize the performance of the network by justifying the use of the largest possible bandwidth capacities.

To accomplish this goal, the user may resort to WAN network consolidation, which amounts to finding ways to effectively transport all inter-site traffic over the most appropriate wide area service. The user, when it's time to consider technicalities, will address issues of the sort listed in the sidebar entitled Requirements For Consolidation Of Wide-Area Networks.

APPLICATION TRAFFIC CHARACTERISTICS
WAN network consolidation is heavily influenced by application traffic characteristics of existing traffic and by the areas of growth. A typical traffic profile is shown in Table 1. (Note that in this table the large growth area is LAN data, and that the great unknown is the deployment of desktop multimedia in its various forms.)

In addition to seeing how applications contribute to overall traffic demands, network managers must pay attention to the particular needs of different applications. The challenge is to allocate network resources appropriately, accounting for applications characteristics, responding with the necessary capabilities.

Delivering Comprehensive Class Of Service Support
Comprehensive COS support to meet the diverse needs of applications is a critical area in comparing the architectural alternatives for enterprise network consolidation. Delivery of comprehensive COS support starts with the ability of the application to implicitly or explicitly signal what it needs, which is complemented by the network's ability to map application needs into WAN COS requirements.

Most applications cannot signal their COS needs. However, this will change, at least in the IP space, where we see the deployment of signaling protocols such as RSVP (the IETF-defined Resource reSerVation Protocol). So implicit signaling is done by assigning ports to a particular application (for example, a room video conferencing system or a voice switching system) or by analyzing the protocol header (for example, voice over IP uses the Real Time Protocol, RTP). Policy considerations will dictate acceptance or rejection of the request based on network resources.

Controlling Latency And Delay Variation
For real-time applications, it is very important to control latency across the network as well as the delay variation across the network. This is done via node-level and network-wide mechanisms required to meet needs of near-real-time and real-time applications. A very critical area is the control of latency over slower access lines, since much higher capacity can be engineered in the backbone of the network.

Reserving Network Bandwidth
A third capability is particularly important for applications that require minimum bandwidth guarantees across the network. Bandwidth reservation mechanisms are required to meet these needs and to provide a basis of chargeback to end users. At the same time, network resources have to be allocated to individual users, which can complicate total dynamic bandwidth operation.

In order to manage network resource allocations, COS-based routing is a necessity to route the traffic across the network recognizing the COS requirements and availability of network resources. This requires that an end-to-end perspective be maintained across the network. Maintaining such a perspective can be difficult, given that traffic patterns change continuously.

ARCHITECTURAL CHOICES
The leading WAN architectural options for enterprise network consolidation include routers, broadband ATM switches, and enterprise network switches. Each of these is discussed in the subsections below.

Routers
While routers were originally designed to route multiple protocols such as IP and IPX, they are increasingly being positioned by vendors for multimedia applications. Enterprise users want to understand the opportunities and limitations in leveraging their installed base of routers to support various forms of multimedia traffic.

The reality is that routers don't support circuit emulation and native ATM switching. However, routers have recently gained limited integrated voice support. Further, more comprehensive external voice-over-IP gateways are being deployed. Under controlled conditions, such as over lightly loaded intranet links, these gateways can deliver quality voice. The challenge is to scale this solution across the entire intranet and support a broad mix of traffic types under all sorts of conditions.

Delivering consistent performance for connection-oriented, real-time traffic across connectionless networks poses a few problems. While today's routers can provide some delay controls through various queuing schemes, delay variation cannot be easily controlled in a router-based architecture. In addition, bandwidth reservation in a router network is soft, in the sense that no firm guarantees over an extended period of time can be provided in a connectionless environment, this being exasperated by the lack of COS-sensitive routing. While router networks will evolve to better meet the needs of voice transport and IP-based multimedia applications, they are currently a poor fit for general enterprise network consolidation.

Broadband ATM Switches
Broadband ATM switches are optimized for broadband ATM switching with non-ATM adaptation integrated into the devices or supported by external muxes. The most common deployment is in a multi-campus environment interconnected by T3 or even OC12 ATM over dedicated fiber or public ATM. Adaptation is provided via the LAN Emulation (LANE) standard and via ATM Adaptation Layer 1 (AAL1) circuit emulation for voice, video, and circuits.

Not surprisingly, broadband ATM switches are designed to meet COS requirements of all sorts of traffic. Delivering consistent performance for connection-oriented, real-time traffic is what ATM is all about. In fact, ATM supports constant bit rate and variable bit rate, both real-time and non-real-time, classes of services. Because of the connection-oriented nature of ATM, the fragmentation of all traffic types into short, fixed-length cells, and the support of COS-sensitive routing, delay variation can be strictly controlled in an ATM architecture.

In addition, bandwidth reservation is hard, in the sense that firm guarantees over an extended period of time can be provided. However, due to the requirement to support T3 wide area connectivity, this architecture is a niche solution that is applicable to a few very large customers (for example, certain federal government agencies) who can afford broadband ATM.

Enterprise Network Switches
Constituting a well-established product class, enterprise network switches were specifically designed to meet wide area consolidation needs, from narrowband to broadband speeds. An enterprise network switch provides multimedia consolidation of legacy and LAN data, as well as circuit-based and ATM-based traffic, and integrates switching and routing to provide low-latency, high-performance networking.

An enterprise network switch provides extensive interface support to consolidate in-building traffic over multiple carrier services and is based on open standards. It combines ATM switching, routing, COS support, and adaptation into a single device.

Enterprise network switches are a proven architectural approach with products available from multiple vendors. They have been deployed in hundreds of enterprise networks, some consisting of hundreds of sites, in virtually every industry segment from finance and healthcare to government and utilities. The business case for an architecture based on enterprise network switches is typically in the 12-18 month range.

CONCLUSION
While broadband ATM switch solutions have their place, enterprise network switches are the basis of a more general-purpose enterprise WAN consolidation architecture, allowing users to optimize the performance of their enterprise networks while simplifying their operations environments, managing their capital investments, and minimizing their recurring costs.

The next step is obvious. Enterprise users have a low-risk architectural solution in enterprise network switch solutions that provide an integrated multimedia switched infrastructure, an infrastructure that meets the requirements for responsiveness to application needs, reliability, security, and cost control. If T1 muxes are slowing down the business of the enterprise, then enterprise network switches can do quite the opposite: They can power the enterprise.

Tony Rybczynski is director of Strategic Marketing and Technologies for Nortel's Enterprise Solutions business unit, which delivers high-performance data networks globally. This business unit creates new alternatives to increasingly complex data network infrastructures, and offers them through direct and indirect sales channels. For more information, visit the company's Web site at www.nortelnetworks.com. E-mail questions or comments to the author at tonyryb@nortel.com.

• In-building interface flexibility and adaptation onto an integrated networking architecture. Three classes of traffic that need to be supported include bit/circuit-based traffic (from sources such as PBXs and video conferencing systems, as well as from legacy synchronous data); frame-based traffic (from LAN-attached workstations and servers); and cell-based traffic (from sources such as ATM-attached devices and servers).
• Carrier interface flexibility and choice (leased line, frame relay, and ATM at speeds from 56 kbit/s to 155 Mbit/s). Traditional leased line services are being replaced by virtual circuit services based on frame relay, at speeds up to T3, and increasingly on ATM, at T1 speeds (up to OC3 and OC12).
• High-performance, low-latency connectivity for the high-growth inter-LAN traffic.
• Dynamic bandwidth on demand for bursty traffic.
• Class of Service (COS) support for circuit emulation, mission-critical data, voice, video, and multimedia.
• Open, standards-based architecture.
• Consolidated network management.