Posted: 09/15/1999
Datacom Positioning: Preparing for the Decathlon
Carriers Walk Tightrope Between Current and Future
Infrastructure
By Peter Lambert
Shifting fundamentals in data communications increasingly require that every carrier makes itself the equivalent of a decathlon athlete. Whether entering the playing field with legacy voice infrastructure, legacy data infrastructure, both types of infrastructure, or a clean slate, players must be fast in the short haul, durable for the long haul, and agile enough to jump higher as customers and competitors raise the bar.
No provider can stand pat. Legacy-burdened carriers must squeeze more life from old technology while simultaneously investing in immature, bleeding-edge innovations. Local incumbents, such as Bell Atlantic Corp., N. Y., are building new access and backbone infrastructures to compete in packet networking while gradually migrating their core infrastructures from circuit- to packet-based technologies. Long distance carriers such as AT&T have already vowed to buy no more old, circuit-based technology and are instead acquiring and building packet networks to extend the range of frame, cell and Internet protocol (IP) services they can offer.
While new carriers shoulder no legacy burdens, they do face limited windows of time in which to establish market share. Carriers cannot wait 12 or 18 months for new technology to become fully baked--and never has so much innovation created so much half-baked, but highly promising, technologies.
To walk this tightrope between present and future, carriers new and old are spending more time assessing new gear, and they are getting help from vendors that offer stepping stones between old and new technologies.
For example, one-year-old competitive local exchange carrier (CLEC) New England Voice and Data LLC (NEVD) of Marlboro, Mass., agreed on July 8 to purchase standard equipment, including 5ESS Class 5 local telephony switches, FT-2000 synchronous optical network (SONET) gear and TNT remote access servers from Murray Hill, N.J.-based Lucent Technologies Inc. Yet, the same three-year, $30 million contract also calls for Lucent to deliver much newer packet-switched technologies, including AnyMedia multiservice access switches, digital subscriber line (DSL) equipment, Connect-Reach integrated access devices (IADs) and, eventually, server-based 7RE soft-switches, the last of which will eventually replace the 5ESS.
"The strategy is to build a network that works now and that opens paths to next-generation services," says NEVD Chief Operating Officer Rob Shanahan. "Lucent has plans for what happens to the 5E switch when they've fully developed the 7RE softswitch; a year from now, we'll look at the 7RE."
Convergence Economics
If their methods differ, virtually all vendors and carriers agree on one common goal: to replace the economics of multiple networks for multiple services with the economics of a single network for all services via a convergence protocol, such as IP or asynchronous transfer mode (ATM).
As of mid-1999, that ultimate goal remains out of reach. Marketplace realities--including customer investments in legacy access equipment--continue to require that providers offer a mix of private data and Internet-based data access technologies, including frame relay, ATM and IP. Consequently, even new carriers must continue to manage complex networks.
However, service providers are demanding their key vendors provide migration paths to simplify their networks to achieve convergence economics.
NEVD, for example, founded in mid-1998 by cable industry pioneer Robert Fanch, is committed to owning, and so controlling the development of, facilities for all its voice and data services. With collocation presence in 53 Bell Atlantic central offices (COs) so far, NEVD is purchasing its own 5ESS circuit switches. Through Marlboro-based subsidiary Fibernet LLC, NEVD also owns fiber facilities in Charleston, W.Va., John-stown, Pa., and Oxford, Mich. (much of it acquired from cable companies), and it intends to own fiber linking COs, Internet points of presence (PoPs), voice switches and backbone network in all areas it serves.
"Where we have access to fiber, we're installing our own switches and DSL and dial facilities, so we're building a facilities-based position in Internet, private data, DSL, and local and long distance telephony," says Shanahan. "By owning and controlling the facilities, we have the flexibility to tailor specific services for specific customers, as well as the ability to manage performance end to end."
In respect to Internet facilities, NEVD earlier this year acquired Boston-based Internet service provider (ISP) OEM.net. Targeting tier 2 and tier 3 markets--beginning with Nashua, N.H., Providence, R.I., and Worcester, Mass.--NEVD is looking for additional ISPs to acquire and integrate into its regional networks.
Finally, to drive its costs below competitors, NEVD is unwilling to build separate backbones for its voice and data networks. Instead, it plans to install cell relay core switches to interconnect all its voice switches and ISP PoPs via a single ATM network.
"We have to look at segments of time, and while we don't want to be bleeding edge, we also don't want to wait for new technology and risk missing current opportunities," Shanahan says. "These tier 2 and [tier] 3 markets will support only so many providers."
Consequently, while it seeks market share with proven access and switching gear, Shanahan says, NEVD is "very conscious of needing not to be stuck with legacy gear, so we've been spending a lot of time with vendors."
Early Adoption
To establish convergence economics will require willingness to stretch new muscles, mainly in the form of risking early adoption of cutting-edge convergence technologies.
Unlike many newcomers to local access competition, AT&T and MCI WorldCom Inc. have the money and clout to take such risks. Both have become early testers of prestandard multiprotocol label switching (MPLS) technologies designed to add ATM-like traffic engineering capabilities to IP networking.
MPLS promises to revolutionize IP networking. While standard send-and-pray routing may send packets from a single data session across dozens of different router paths at various speeds, MPLS-based routing labels packets as belonging to a single flow, then engineers a single router path for all packets in that flow. By bringing connection-oriented, circuit-like virtual paths to the connectionless routing process, MPLS assures more consistent transmission quality.
Last January, AT&T announced deployment of MPLS capabilities in its popular frame relay and ATM networks, which originate in 41 U.S. markets served by subsidiary TCG and then travel over AT&T's global backbone. "With standard frame relay, you have to pre-engineer a permanent virtual circuit end to end through the network," says Jim Daugherty, general manager of AT&T's data services marketing. "Now, with MPLS, you don't have to pre-engineer the connections." That means making IP a more complete convergence protocol, less reliant on intelligence residing with other underlying protocols like frame relay and ATM.
Another key area of convergence technology risk lies with integrated access devices (IADs) that enable corporations to access multiple voice, private data and Internet backbone networks over a single local access network. For example, like Kansas City, Mo.-based Sprint Corp.'s Integrated On-Demand Network (ION) service, AT&T's Integrated Network Connection service now converts all data packets and time division multiplexed (TDM) voice traffic to ATM cells at the customer premises for transmission over a single local ATM network.
"Where customers access our network via TDM, the bulk of traffic is becoming data. To accommodate this change, this service takes packets into the network using ATM, which can achieve scale and quality of service (QoS) controls," Daugherty says.
By replacing TDM access with ATM access, Daugherty says, AT&T achieves two key goals: improved performance through more efficient, dynamic bandwidth allocation and improved costs by allowing AT&T to bypass local access charges levied by local TDM carriers.
Dense wavelength-division multiplexing (DWDM) also promises to provide similar benefits by enabling low-cost, dynamic allocation of bandwidth via protocol-independent wavelengths, and vendors like San Jose, Calif.-based Artis Software Corp. are trying to mitigate the risk of experimentation by delivering engineering planning and simulation tools for new optical and networking technologies.
"With so much innovation," says Artis Software President and CEO Marco Baldassari, "you wouldn't be able to live without simulation."
Picking the Low Fruit
The sooner such new technologies can help pay for new data infrastructure, the sooner a carrier can build stamina for the datacom marathon. One such early revenue maker lies with IP virtual private networks (IP VPNs), particularly to meet the burgeoning demand for mobile and home worker dial-up remote access into corporate networks.
To meet this demand, AT&T, Bell Atlantic and many other carriers also are supporting IP security (IPsec) encryption and encapsulation technology based on the Internet Engineering Task Force's (IETF's) emerging IP security standards. Combined with MPLS, IPsec can enable corporations to carve out low-cost, secure, private tunnels across the public Internet for remote access, corporate site-to-site intranets, business-to-business extranet commerce and IP telephony services.
"Remote access is the low hanging IP VPN fruit," says Bill Jefferis, senior product manager, VPN services for Bell Atlantic, which last June launched a managed VPN service that includes network design, procurement of wide-area infrastructure, installation and management of VPN customer equipment, deployment of IPsec client software, distribution of digital certificates (another layer of controlled access security) and service level agreement (SLA) guarantees and monitoring.
"We're in six pilots of 25 to 50 users now, and once it's proved out over about 60 days, it becomes a tremendous annuity service that can continually add hundreds and thousands of users, as well as lay a foundation for site-to-site, extranet, firewall, enterprise application hosting and other managed services," Jefferis says. "All of that can be built on top of IP VPNs."
Because remote access generally requires wide geographical reach, carriers also must invest in dial-up PoP infrastructure. To achieve this quickly, AT&T purchased IBM Corp.'s global IP network with more than 2,000 local dial PoPs worldwide, and Bell Atlantic is building its own PoPs in its Virginia to Maine service region while partnering with GTE Internet Solutions, Irving, Texas, to gain dial-up reach nationwide.
Also experimenting with packet-to-circuit gateway technology, AT&T has begun to roll out business IP telephony, primarily for international price arbitrage, and is watching and testing development of gateway control protocols and platforms for development of packet-based call control applications.
"We're not buying any more Class 4 toll switches after this year. We're building IP backbones as large as anyone else's, and we'll arrive at packet-based telephony as quickly as anyone else," Daugherty says. "But this all will evolve slowly because it will take time to match both the features and reliability of the public switched telephone network (PSTN)."
Bucks, Not Bits
Indeed, while it may seem that pure invention is driving fundamental shifts in datacom infrastructure, "ultimately, cost is what drives customers to change in any substantial way how they do things," says Daugherty.
Because they use shared, public Internet infrastructure, IP VPNs promise to provide lower costs per customer. However, for dedicated data connections, few carriers expect IP VPNs to replace frame relay overnight, in part because of investments in frame relay equipment and in part because shared, quasi-public frame relay networks also are virtually ubiquitous.
"Frame relay and ATM are shared transport too, so the costs really are somewhat similar to shared Internet infrastructure," Daugherty says. "We still see strong growth in frame relay and very strong growth in ATM, driven by data."
Also bullish on ATM backbone equipment, Rochester, N.Y.-based Frontier Communications Corp. is nevertheless convinced that IP networking will radically improve service provider efficiency and economics. Consequently, earlier this summer, Frontier committed its future to IP convergence, agreeing to purchase packet telephony gear that is thoroughly bleeding edge.
That gear includes Lucent's local telephony softswitch and San Jose, Calif.-based Sonus Networks Inc.'s large-scale Open Services Switch, the latter of which Frontier will use to interconnect its old circuit-switched trunking networks with its newer packet-switched networks. All of this will run over a fiber network driven by cutting-edge DWDM optical networking equipment.
Frontier plans to launch IP telephony in 15 to 20 U.S. cities by mid-2000. By year 2002, the company plans to unify access for all its traffic--voice included--over IP, and to extend that network worldwide through its prospective new owner, Global Crossing Development Ltd., Bermuda.
Like New England Voice & Data, Frontier will use emerging DSL technology to expand its access networks and relatively mature ATM technology as a bridge from pure circuit-switched infrastructure to pure IP. "We have two very strategic partners: Cisco [Systems Inc., San Jose, Calif.] in our backbone network and Lucent for voice over IP (VoIP) and gateways, which initially we'll run over our ATM network," says Allan Van Buhler, vice president of new product development for Frontier.
Assuming that emerging IP QoS standards and protocols mature and are perfected, "we think we can be an early adopter and eventually go to pure IP, but that will be an economic decision, not one forced by the technology," he says.
Whether via IP and ATM or IP alone, he adds, the economics of packet switching are
extremely compelling, even in terms of data-based telephony alone. Frontier estimates its
adoption of a distributed, server-based IP telephony switching infrastructure will save 50
percent in port costs,
and 50 percent to 75 percent in collocation costs, compared to centralized,
mainframe-based Class 5 and Class 4 circuit switches. Van Buhler projects packet switching
also will yield eight to 10 times more efficient bandwidth usage than dedicated circuits
can offer, and will require a smaller staff to manage.
"We can see these savings in voice alone, and in converged networks handling both voice and data, the economics get even better," says Van Buhler, who adds that migrating to a single network for all services would reduce two network management systems to one, thereby yielding savings in people and training.
Risky as all the new technology experimentation appears, most agree that experiments guided by the goal of improved efficiency and economics can't lose, given patience.
"The new services like site-to-site VPNs are a little harder to prove than remote access VPNs, particularly where frame relay is working well," says Jefferis. "Over the first few years, the cost may be a wash because of dollars already sunk in frame relay, but after those contracts end, the benefits of unloading the customer's frame relay management and ownership burden will be easy to demonstrate."
