networking? Or will it collapse under the weight of its own success, as cynical experts have suggested? To understand the technological challenges of establishing any-to-any digital networks, let's look at these hurdles one by one.
Puzzle Pieces
What technologies have to converge to make any-to-any networking a reality? Start from the top with advanced backbone platforms like synchronous optical network (SONET), asynchronous transfer mode (ATM), and Switched Multi-megabit Data Service (SMDS). Add to this relative newcomers to network access like ISDN, asymmetrical digital subscriber line (ADSL), and frame relay. Traditional technologies are part of the convergence puzzle, too, so you have to factor in networking standbys like X.25, T1, Systems Network Architecture (SNA), and dial-up networks over private telephone lines.
Like the Wizard of Oz, who ruled the la
nd as that man behind the curtain, data networking providers -- from telcos to long-distance carriers, Internet service providers (ISPs), cable TV operators, and systems integrators -- must work under the covers to solve the complexities of interconnecting numerous backbone and access technologies. Corporations working out an any-to-any networking plan will find as many strategies as there are data networking providers, which means that the ultimate solution will probably be a mix of technologies. What's more, the right mix for businesses will probably be different than what's used in homes. In the latter, everything from plain old telephone service, Basic Rate Interface (BRI) ISDN, ADSL, hybrid fiber coax (HFC), cable modems, and switched digital video (SDV) will play roles.
In the business world, "clear channel" TCP/IP using 56-Kbps or T1 facilities, frame relay, and Primary Rate Interface (PRI) ISDN will initially dominate, says Dave Schriftgiesser, broadband networking director for Lucent Technologi
es, a designer of communications systems and software (and formerly known as AT&T Network Systems). "As ADSL and cable modems come on-line, they will overwhelm frame relay's capacities and we'll be migrating to ATM," he adds.
For corporate customers, "full service" is the message being heard from data-networking providers. The list of offerings includes voice, video, and data services via wireline telephony, personal communications services, videoconferencing, cable TV networks, digital near video on demand (NVOD), and high-speed access to the Internet. In some cases, these services will be regional; in others, they will be national.
For example, Bell Atlantic, which is in the process of merging with fellow Baby Bell Nynex, offers All@once, a package of data networking products that use SMDS, FDDI, ATM/cell relay, frame relay, and multiple dedicated private connections. Bell Atlantic also plans to eventually deploy a network that has a switched digital video backbone with the final thousand feet c
onsisting of ADSL over twisted-pair connections. "Our long-term vision is to deploy a complete end-to-end ATM solution," according to Bill Lawrence, vice president of network systems engineering for Bell Atlantic. "However, ATM is still in its infancy, and we don't have a big enough fiber backbone in place to support it at this point," he says.
For the near term, Bell Atlantic is concentrating on "in region" data networking solutions, where customers typically have frame relay and/or SMDS at the ends of their networks. "ISDN is more problematic, but we pop out into an SMDS cloud and then we convert as needed to other access technologies. For the most part, we try to persuade our customers to direct-connect to our SMDS network," Lawrence says.
The U.S. Federal Communications Commission is working on a variety of regulations that will affect how RBOCs provide data networking services. "Right now, we are still required to pass our data transport to long-distance carriers when we cross LATA [local acc
ess and transport area] lines," Lawrence says. "In the future, we expect the FCC to drop these restrictions, and when they do, we'll revise our data networking model."
Life at the Application Layer
How can you protect your investment in LANs? MCI and other carriers hope to provide services that help internetwork legacy systems, like SNA networks, and migrate them from a host-based design to a client/server architecture.
Any experienced MIS director appreciates the pain involved in installing a LAN and the affects it has on end-user applications. The problem becomes magnified as more corporate LANs are interconnected to WANs and metropolitan-area networks (MANs). "From a wide-area networking perspective, we'll reach nirvana when our network evolves to a campus-like environment and applications
are transparent
," says Steve Starliper, vice president of marketing for !Nterprise Networking Services, a division of US West Communications and a provider of LAN inter
connection and ATM backbone services.
The !Nterprise strategy is to move beyond providing customers with a "dumb" bit pipe. Many telcos provide services between Layers One through Three (the physical, data link, and network layers, respectively, of the Open Systems Interconnection reference model). !Nterprise creators are focusing further up the protocol stack to Layer Seven, where applications for such things as e-mail, file transfers, and transactions reside. !Nteract could achieve applications transparency by bundling bandwidth and application offerings. This is a new strategy for data-networking providers, who typically assume customers will provide their own applications. !Nteract uses Microsoft's Windows NT and BackOffice Server suite as its operating platform and incorporates Netscape's Navigator for "plug-in" access to the Internet. In addition, the service uses Cisco Systems' internetworking platform and protocol support. Other value-added services include Web hosting, groupware via Lotus Notes
, local content, fax, forms management, vertical market applications, and transaction processing.
The strategy sounds good, but data networking providers haven't proven they can deliver on the promise. This approach requires providers to do some software engineering. Nevertheless, some people believe that an !Nterprise-like plan is the best hope for any-to-any digital data networking. "The carriers [must] break down walls and move up the value chain toward providing systems integration services," says Don Cleavinger, chief technologist for EDS's communications group, a data networking integrator. "Any carriers that stay in the commodity business of providing bandwidth and cheaper T1 service will be missing the boat. Others that move up the chain will have a role to play [in digital internetworking]."
MCI recognizes the potential of this strategy but doesn't implement it to the extent that !Nterprise advocates. MCI services handle Layer Two protocol conversions, but the company expects customers to
take care of Layer Three conversions (e.g., IP to IPX). To make this transition, MCI in part will use routers to connect customers to its SMDS or frame relay services. But the company cautions that merging legacy systems into a client/server architecture isn't for those who need instant gratification. "On a big network, executing a transformation plan can take up to a year, especially if Layer Three internetworking is required," says Steve Tabaska, MCI vice president for data services engineering.
MCI expects to move from today's separate networks to an integrated router scheme that supports multiple protocols to achieve internetworking. "In effect, we're trying to create an integrated router from a LAN environment, and then evolve the LAN into an integrated WAN router environment," Tabaska says. For example, MCI offers X.25, ISDN, frame relay, SMDS, IP, and ATM services. Some of these can be internetworked, while others cannot. "ISDN is really a Layer One service; X.25, frame relay, SMDS, and ATM are
Layer Two; and IP is Layer Three," Tabaska says.
Currently, MCI supports seven combinations of internetworking for its services, including dial-up to X.25; ISDN to X.25, frame relay, and IP; and frame relay to X.25, IP, and ATM. MCI does not support dial-up and ISDN access connections to SMDS and ATM, or frame relay to SMDS, among others. "Evolving a complete internetworking plan for MCI's offerings will take two years to put in place," Tabaska says.
Sprint's answer to internetworking is ATM to IP and ATM to frame relay, according to Cathy Gadecki, group manager for ATM services. The company plans to bring on-line next year a pure ATM backbone for all frame relay and IP end points. "We'll be pushing out the frame relay backbone to the network edges and then phasing out our frame relay backbone," Gadecki says.
In its current circuit-switched network, Sprint uses either ISDN or Switched 56 services to provide access to its frame relay network. "In the near future, ISDN will become an access te
chnology as well as a backbone," according to Gadecki.
Sprint has no intention of missing the "full service" boat. From a transport point of view, cost effective and easy-to-use ISDN, ATM, frame relay, and IP will be key enablers. "Sprint plans to complement our transport capabilities by providing managed network services, bundling in CPE [customer premises equipment] offerings, providing IP security services, as well as other Internet and intranet services," says Gadecki.
While telcos want you to look no further than their product brochures for ways to integrate applications, some competitors are building software tools to help you turn convergence cacophony into harmony. Lucent recently announced a software platform,
called Inferno
, that could provide applications transparency on any-to-any data networks. "It's the client/server multimedia equivalent of dial tone," Schriftgiesser says. "Inferno relates to the signaling networks of today and will provide signaling for the tra
nsaction-oriented point-and-click world as well."
Inferno consists of a network operating system; communications protocols, called Styx; a network API; a programming language called Limbo; and a virtual machine called Dis. Lucent will license Inferno to service providers, telephone and cable companies, applications developers, and content providers. Inferno may even find a home in Internet terminals, hand-held devices, advanced telephones, and set-top boxes.
The Network Is the Home
On the residential side of digital networking, Time Warner's estimated $50 million interactive cable TV system in Orlando, Florida, called the Full Service Network (FSN), has become one of the most notorious examples of voice, video, and data services. Using a 1-GHz HFC network, FSN incorporated an ATM-to-the-home scheme that required a massive systems integration effort and more lines of code than the first Apollo moon shot. The project has logged several delays and substantial cost overruns. By earl
y 1996, the service had more than 4000 subscribers, but the estimated cost of the network was more than $10,000 per subscriber.
Because of these mixed results, other companies are taking a more modest approach to the residential market.
@Home,
a provider of high-speed Internet access and on-line content, uses cable modems exclusively and departs from the conventional wisdom of offering a wide variety of access methods to the Internet. By 1998, @Home (cofounded by cable giant TCI) hopes to reach one million subscribers using this strategy.
The @Home network (in parts of California, Connecticut, and Illinois) will use a conventional TCP/IP platform. As it expands to other geographical areas, @Home will continue to support end-to-end multicasting using standard Internet protocols. The company also will deploy its own ATM backbone network by leasing or purchasing access from third parties across the country. This approach means @Home can keep its technology focus tight: The majori
ty of internetworking issues will center around cable modems and the cable plant itself.
While @Home and others believe in the cable TV network as a way to deliver data, PSINet, an ISP with more than 320 points of presence (POPs) worldwide, is less optimistic about that platform. Over the last several years, PSINet has worked with Continental Cablevision in Cambridge, Massachusetts, to test high-speed Internet access via cable modems. The trial proves that cable modem hardware works, PSINet says, but Internet connections over the telephone network are still better than those over cable networks. "What doesn't work is overlaying cable modem systems on the existing cable plant," says PSINet CEO William Schrader. "There are many problems with upgrading this plant to provide two-way capabilities and to eliminate noise that makes data connections unreliable. We believe it is better, cheaper, and faster to use the telephone plant."
Internet Dominance
At the heart of any-to-any interne
tworking is a debate over backbones. On one side are those people betting network backbones of the future will be based on an ATM switched platform. Others see the Internet's IP-routed platform providing a glimpse of the future.
"The common denominator for any-to-any digital data networking is the Internet, and the common denominator for the Internet is IP," says Ron Vidal, MFS Communications' vice president of new ventures. "In fact, true any-to-any data networking is fostered by the Internet because IP can reside in a client, server, or any other device."
MFS, a competitive access provider, jumped on the Internet bandwagon when it merged with Internet service provider UUNET Technologies earlier this year. The combined company is the only ISP (out of about 2500 in the U.S.) to own or control fiber-loop, intercity, and underseas facilities in the U.S. and parts of Europe. This international platform interconnects 543 Internet POPs, including 288 outside the U.S.
MFS is bullish on the Interne
t in part because it is much cheaper than alternatives. For example, the cost to deliver a 42-page document between New York and Tokyo via the Internet can be as low as 2.8 cents -- versus $28.83 over an AT&T phone connection during peak times. The Internet is more cost-effective than circuit-switched networks because optical technology has made transmission capacity relatively inexpensive and because IP packet switching uses transmission capacity more efficiently than circuit switching does.
"Also, I can make a good case for analog-to-IP connectivity when we're talking about Internet phones," Vidal says. "It's conceivable that people will even start designing PBXes to connect to packet-switched IP networks. Maybe the Internet starts looking more like a phone network at the physical routing level."
MFS plans to offer Internet services ranging from metro-area exchange connections to local and national backbone access. "In our ÔInternet hotel' concept, a customer can colocate a server at our central o
ffice and obtain the capability to increase the speed of a 56-Kbps connection to a T1 connection in a matter of hours," Vidal explains. "In contrast, the telcos are not willing to do this."
Businesses can set up even more sophisticated networks on intranets than on the Internet because intranets typically have more bandwidth available than dial-up connections, according to Vidal. Therefore, he believes, intranets can handle larger files and more complex graphics at higher speeds. "Ultimately, the LAN-to-LAN connectivity business will become an intranet business," Vidal predicts.
PSINet is also an unabashed booster of the Internet as the any-to-any networking platform. "The Internet scales well and it's ubiquitous" says Mitchell Levinn, PSINet's vice president of network operations.
"Ninety percent of all computer programmers on the face of the earth are working on internetworking applications," says PSINet CEO William Schrader. "Data movement of all types is being carried over TCP/IP networks.
In the future, we can look forward to TCP/IP riding on an ATM network or SONET OC-12. In five years, we'll see TCP/IP packets riding over ATM to the desktop."
However, the Net's popularity is why other companies aren't putting all their packets in one basket. Internet traffic is getting so heavy that some people are predicting that the Net will begin to crash under its own weight later this year. While not going this far, Lucent's Schriftgiesser acknowledges there are "brownouts" on the Internet. He contends that the costs and benefits of providing additional capacity are so dispersed that people aren't motivated to really do anything about it. "The current telecommunications infrastructure is built around voice calls that require relatively short [connection] times," Schriftgiesser says. "Meanwhile, heavy Internet usage and browsing involve relatively long [connection] times and are thus loading down the telecommunications infrastructure. This leaves carriers trying to position for the future, while a
ttempting to fix problems within their current systems."
Digital Future
All the empirical evidence points to an evolving telecom world in which data communications internetworking and any-to-any connectivity are essential within companies and among enterprise networks. While data networking carriers and equipment providers acknowledge the importance of any-to-any connectivity and promote their own visions for making it happen, few of them will admit that providing it to their customers is easier said than done.
Any-to-any digital data networking will become ubiquitous, global, and a vital part of communications. However, convergence may take the rest of the decade, cost more time and money, and cause more headaches than even savvy data communications professionals anticipate. But when we look back from the perspective of 2006, most of the challenges discussed here will probably have been met, and we'll remember convergence's technical hurdles as nothing more than growing pains.
Where to Find
Lucent Technologies
Murray Hill, NJ
Phone: (888) 458-2368
Internet:
http://www.lucent.com
Convergence Strategies
