December 1999

Millenial Vision: Lifting Our Gaze Past The Smoke And Mirrors

BY TONY RYBCZYNSKI

Alternately dazzled and disappointed by hype (and distracted by Y2K), those of us who are responsible for enterprise networks could be forgiven were we to confess that the new millennium, now so close, occasioned at least as much confusion as inspiration. And yet, while inspiration often eludes us, we should make the effort to define a networking vision for the enterprise. If we don’t, we could miss the very real opportunities within our reach.

Many of these opportunities are contingent on instances of convergence:

• We are already witnessing greater connectivity and a proliferation of new applications — thanks to the unification of telephony and data.
• Connectivity and applications will progress yet further in the wireless era — as broadband wireless mobilizes Internet users, fomenting yet another application revolution.
• The Internet is becoming the engine for e-business — now that barriers between enterprise and carrier networks are collapsing.
• LAN/MAN/WAN networking is becoming seamless — thanks to optical networks.

In all of these examples, convergence is a key enabler. Indeed, opportunities are multiplying as the world accelerates its pace towards convergence onto IP and the Internet. At the same time, however, our vision is obscured by a thickening smoke screen of myths. These myths arise from various sources. Some myths originate with vendors, be they startups or established companies. Other myths start with pundits, analysts, and the press, as they all scramble for “mindshare.”

Fortunately, it isn’t all that difficult to see through the smoke. With a little effort, we can distinguish between fact and fiction, allowing us to grasp the opportunities occasioned by convergence, and to avoid making costly mistakes, whether through ill-considered actions or even inaction.

It’s all about making sound decisions. And sound decisions are based on facts. So, in this column, I will discuss a few myths, the better to dispel them, and reveal facts that might otherwise remain hidden.

MYTH #1: Bandwidth Will Be Free (Not In My Lifetime!)
The anticipation that bandwidth will be free is rooted in the explosive growth in the number of bits that can be put on a single fiber, which is now in the multi-terabit/sec range (that’s twelve zeros or a million million bit/s). We could even say that bandwidth is growing at a faster rate than that of Moore’s Law for processors. So, should we be hopeful?

Certainly, the cost per bit for 10-Gbit/s pipes is coming down, but you need a lot of traffic aggregation to fill these pipes. Also, just as we develop applications that consume more and more MIPS and memory bytes, we will roll out applications that will gobble up ever greater amounts of bandwidth. In addition, there is the cost of terminating high-capacity fiber and switching all those packets.

So, even in campus networks, in which the installed cost of fiber capacity is relatively low (compared to the WAN), the end-to-end cost of capacity is still substantial. In addition, planning for prospective traffic growth remains difficult. (How many fast Ethernet users can swamp a gigabit Ethernet pipe? Fewer than 10!) And supporting this traffic growth while acquiring (or maintaining) the ability to overcome failure conditions is a real challenge.

And then there is the tariff question, since the business case for a carrier to offer free bandwidth is poor at best. In addition to selling raw bits, carriers are also looking at premium-priced managed service, IP quality of service (QoS) mechanisms, and virtual private networks (VPNs) to enhance their business models.

What to do? Take advantage of bandwidth to improve application performance and minimize traffic engineering; design for highest performance for the dollar, as opposed to squeaking by with the lowest bandwidth that will meet your present performance needs.

Traffic will grow to fill the pipes, so scalable switching capacity is a must across the board. You should also start going down the path of defining some policies in your network that will identify which users and applications merit preferential treatment in the case of congestion or failure. To begin delivering on these policies, you may want to introduce basic IP QoS capabilities in critical areas (for example, WAN links).

So, will bandwidth be free? I expect that a good chunk (up to 30 to 40 percent) of your communications budget will continue to go for bandwidth and WAN connectivity.

MYTH #2: Voice Over IP Is Free (No! But Freedom To Choose Is Critical)
Integrating voice traffic onto your data network can save you money. And if you ignore bandwidth, equipment, and people costs, you can say voice can be free. However, Jim Metzler, a noted consultant, hastens to point out that “anyone who talks about free voice has never run a real network.”

Enterprises have been integrating voice and data on their enterprise WANs for years, initially using TDM multiplexors and more recently running voice-over-packet technologies such as ATM and frame relay. The business justification for moving from channelized to packet-based operation typically cites a payback of 6 to 18 months, particularly when applied over trans-border and off-shore networks.

The introduction of various voice over IP gateway products provides yet another option. These products leverage the same base technologies that have been used on frame relay and ATM. Such technologies include digital signal processors, which may support low-bit-rate voice coding and echo cancellation.

But if IP telephony isn’t about free voice, what is it about? The reality is that the cost of an enterprise communications system is predominantly in the telephone sets and in the application and feature support, including messaging, automatic call distribution, and CTI applications, and not in switching. IP telephony opens up new application opportunities, along with more connectivity options over LANs and IP networks, including the Internet.

Key opportunities in the short term include teleworker voice/data capabilities, collaborative tools (for distributed workgroups), unified messaging, and Internet-enabled call centers. These applications will by necessity be accessible from new IP telephones, as well as from traditional telephone sets linked to PCs through computer telephony interfaces. What should be free is the ability to choose the migration path and rate at which the enterprise leverages these IP telephony developments in line with business objectives.

MYTH #3: Legacy = Scrap (Err, Think Migration Path Instead)
Marketers often generate self-serving slogans, which often confuse users and raise expectations beyond reality. Some of these slogans refer to “Old World” and “New World” technologies. As you might guess, the marketers imply (all right, insist) that investments in Old World technologies should be scrapped today in favor of New World technologies.

The Old World technologies often referenced include older protocols (for example, DECNet, IPX, and SNA), campus ATM/FDDI/token ring, frame relay/ATM WANs, software-based multi-protocol routers, and circuit-switched networks. The New World is based on IP, Ethernet, Internet VPNs, wireless access, and optical backbone networks. No question that these technologies are major enablers of a new era in networking, which will be orders of magnitude more reliable and scalable than today’s networks.

But are all the remnants of the Old World due for the scrap heap? In the Real World, this is not the answer enterprise users can accept unilaterally. They are looking for migration paths and ways to leverage their non-depreciated investments. Let’s look at a number of attributes of Real World networking.

Application Convergence Onto IP
There is no debate that, industry wide, a plethora of new applications are being developed on IP, and that the source of most traffic growth in enterprises is coming from IP. Another enterprise preoccupation, Y2K, has prompted many to redesign existing applications to run on IP.

In spite of these trends, in many enterprises, a significant amount of traffic, much of which is mission-critical, is running on protocols such as SNA. Any suggestion that the industry as a whole should invest literally hundreds of billions of dollars to redesign these Old World applications, so that they may be IP-based, is laughable. It’s not going to happen.

There is good news, however. There are many gateway technologies that allow non-IP applications to run over IP networks. Transporting all applications on IP networks has significant positive total cost of ownership implications by decreasing network management complexity and allowing the enterprise to leverage the power curve of IP-optimized networking, using (for example) routing switches with 10 times the price performance of routers in campus networks.

Campus Convergence Onto Ethernet
There is no question that 10/100 Ethernet has won the desktop from a price perspective, and that Ethernet routing switch products can now deliver the reliability and performance of technologies such as ATM, FDDI, and token ring. At the same time, these other technologies (particularly ATM, in campus backbone networks) represent a large installed base that continues to serve its purposes well.

Enterprise users don’t want to write off this investment, they want to leverage it while taking advantage of migration paths to gigabit Ethernet-based backbones. Vendors are stepping up to this challenge by providing gigabit Ethernet uplinks on their ATM products with reliability provided through mechanisms such as multi-link trunking. Major capacity-driven investments, in ATM campus networks, for example, may hasten a transition to totally Ethernet-based solutions.

Routers As Multi-Protocol Gateways
An explosion in IP traffic. A proliferation of dedicated multi-megabit access systems (including 10/100 to desktops and xDSL/cable modems to the home). Together, they impose ever more severe demands on IP routing systems. We need high-performance, highly reliable, highly scalable, and cost-effective IP routing systems.

As we roll out such systems, we will take IP networking beyond the “best effort” environment associated with the Internet and open up the network for new classes of mission-critical, latency-sensitive applications. Specialized backbone and edge hardware intensive routing switches will displace traditional routers as the workhorses of IP public and enterprise networks. Multi-protocol software-based routers will still play a role in evolving IP networks, but they will be used as multi-protocol server gateways at the edge of the IP network, providing a transition path for the installed base of non-IP application traffic.

New Carrier IP (Layer 3) Services That Complement (But Don’t Replace) Traditional Layer 1 And 2 Services
Some vendors suggest Internet-based VPNs will replace all other enterprise-oriented services, and that the transition will be accomplished within a few years. How realistic are these assertions?

Remote access and extranet VPNs definitely represent important opportunities, since they are enablers for new ways of doing business (with partners, suppliers, and customers). They provide cost-effective, secure alternatives to dedicated dial ports for remote users. With the rapid growth of these capabilities, existing investments in private dial capabilities can be capped. Site-to-site connectivity, however, is another matter.

Even if all enterprise traffic is carried over IP, large carriers will continue to offer a full spectrum of Layer 1 (pipes and wavelengths), Layer 2 (frame relay and ATM virtual circuits), and Layer 3 (IP-based) services with differing price/performance attributes.

A particularly interesting trend is establishing Ethernet as a non-traditional interface to carrier Layer 2 and 3 services. For larger fiber-attached sites, this implementation will include 10-gigabit Ethernet, which is being designed to work in both LAN and WAN environments. It’s the next campus backbone platform, and it provides an effective payload for wavelengths in DWDM (Dense Wave Division Systems) optical systems.

So, IP-based services are a significant new service option offered by carriers, but they don’t constitute a panacea. They won’t replace all other services.

MYTH #4: Circuit Switching Is All But Dead (No, It Will Assume A New Role)
The same arguments that accompanied the demise of the T1 mux as a strategic wide area platform are being used to suggest the demise of circuit switching, the core technology of PBXs and the public voice network. But do these arguments really carry over to circuit switching? Before we answer this question, we should review the differences between the T1 mux and circuit switching.

PBXs are circuit switches that provide connectivity for directly attached end users and are the basis of highly reliable, highly featured telephony services and applications (such as messaging, call centers, and interactive voice response). On the other hand, T1 muxes were nothing more than circuit-based wide area consolidation platforms that accommodated a mix of data, video, and voice traffic. T1 muxes didn’t offer end user services at all.

With these differences in mind, we might see a role yet for circuit switching, even as IP telephony promises to replace the switching fabric, with connectivity to LAN/IP technology enabling a whole new range of applications such as unified messaging and Internet-enabled customer care. The role of PBX common control may evolve. It may become a server on an IP network, while line and trunk-side functionality becomes a gateway to the IP network for the installed base of analog and digital telephone sets, as well as a gateway to the public switched telephone network (PSTN).

In new locations, IP telephony presents opportunities to either distribute the central site PBX functionality to the remote site over an IP network, or establish a relatively autonomous IP telephony systems with gateways to the PSTN and back to the central site.

SEEING BEYOND MYTHS TO APPROACH AN ENTERPRISE VISION
While networking technology can be diverting, the ultimate test of whether any particular technology merits attention is, in a word, relevance. In the case of the enterprise, the relevance test raises questions such as these:

How can networking technology help me compete more successfully? Serve the customer better? Make better use of scarce talent and limited financial resources?
To show the relevance test in action, we can apply it to something familiar, the Internet. Originally — just six years ago — the Internet wasn’t thought of as a business tool. It was an academic tool. Even advertising on the Internet was frowned upon. Then, http/html was invented. Suddenly, we had a worldwide information warehouse. A new surfing language was created, and www.anyname.com appeared everywhere.

We are now witnessing a second surge of innovation and change, as the Internet becomes a transactional marketplace, bringing business objectives to the forefront. An entirely new way of selling to and serving customers is being created as the Internet goes interactive, enriching how people communicate in real time.

The example of the Internet suggests that improving networking technology needn’t be a matter of jumping on the next hype bus or scrapping legacy investments. Rather, we can ask ourselves how we can leverage networking technologies for competitive advantage.

We can even go a step further and insist that the network is the business. Which is to say, we can recognize that the network doesn’t evolve on its own, independent of business objectives. Rather, the network may evolve by taking advantage of one or another network solution, and assume a path more or less compatible with business objectives.

In the case of today’s networks, solutions emphasizing convergence suggest several ways in which business objectives may be achieved. These convergence solutions, within the context of business objectives, contribute to a unified enterprise networking vision.

This vision for the enterprise, when aligned with business priorities, addresses unification (or convergence) at four levels:

1. The Networking Infrastructure (including LAN and WAN, and access by telecommuters, road warriors, suppliers, partners, and of course customers).
2. The Management System (including network, policy, and service level management).
3. The Application Space (including opportunities enabled through technologies such as IP telephony).
4. The People Dimension (including operations and vendor partnering).

This vision, while encouraging a comprehensive approach to enterprise networking, is still but an outline. It may suggest different paths to different enterprises. For example, a given enterprise may plan a unique path allowing it to evolve — in business-driven increments — from its existing environment towards its own vision of a unified network.

The important thing is to have a vision. If you lack a vision, you may wander aimlessly through a fog of competing claims and preposterous promises. If you do have a vision, however, you’re more likely to take decisive action — to start down your own path to where you need to be to compete in the next millennium.

Tony Rybczynski is director of strategic marketing and technologies for Nortel Networks’ Enterprise Solutions unit. This business unit offers a full range of enterprise terminal, workgroup, campus, and wide-area unified networks and applications, through direct and indirect channels. For more information, visit the company’s Web site at www.nortelnetworks.com. E-mail questions or comments to tony.ryb@nortelnetworks.com