20/20: Network Visions for 2002 and Beyond - part 2 of 2
from: telecoms-mag.com
--------------------------------------------------------------------------------
Jay Adelson
Founder and CTO
Equinix
Equinix is in the cat-bird seat as the calendar turns over to a new year. As quickly as the dotcom era arrived, a whirlwind of economic corrections have brought consolidation and cost-consciousness to the Internet. In terms of both the business models and the actual networking technologies, Equinix sits in the middle.
Companies such as Google, MSN, Yahoo and other major content providers have located to Equinix's IBXs (Internet business exchanges) because of cost efficiencies found close to the backbone. According to Equinix's founder and CTO Jay Adelson, these leaders are moving swiftly toward more advanced forms of content peering, which Equinix enables.
"This whole notion of having to pay per packet to get your service to your eyeball-to its destination-isn't necessarily what's going to happen anymore, particularly with the larger players. Content peering is an important trend to watch, and I think its ramifications are significant," Adelson says. "Traditional content services like Reuters and Lexis-Nexis have always been able to charge for access to what they've got." But that may not continue, he says. Adelson expects new content-driven business models to effect the economics of products offered by carriers, ISPs, content distribution networking companies, and even equipment vendors in the next 12 months.
Changes in that direction have already begun. The size of the equipment on the interface is getting bigger, the processing power that sits behind it is getting faster, and the architectures are becoming more consolidated and better organized. "Bringing technologies that were in the hands of carriers into the hands of enterprise will have a radical change on how business is done in this space," Adelson says.
MAN service providers, he says, are already moving away from selling IP bandwidth to selling Ethernet or are going back to selling SONET. "The DWDM explosion is now manifesting itself in appliances and network devices that make it much less expensive for enterprises to integrate into the networks," Adelson says. "It's not necessarily providing bandwidth, but rather managing that bandwidth and the networking components for enterprises that are young to the advanced networking world."
"Because we sit in the middle, we're going to be very focused on optical switching technologies. We're very involved in the IETF and watching how BGP (Border Gateway Protocol) changes to support changes in Internet makeup," he says. Equinix will continue to evaluate some of the latest optical technologies including Calient optical switches, terabit routing from Mazoo, advanced DWDM and passive switching technologies.
-Jared Bazzy
--------------------------------------------------------------------------------
Brian Brewer
Senior Vice President & Chief Marketing Officer
WorldCom
The future of networking may center on IP, but Brian Brewer says businesses are reluctant to spend a lot of money on new types of networks and applications. Because companies aim to run existing applications more cheaply during a struggling economy, growth for services such as VPNs are not as robust as carriers might have expected.
"Especially in this economy and with the reaction to the Sept. 11 events, we've actually seen better growth on the traditional networks than we've seen with IP-related networks," Brewer says. "So we've worked very hard to get away from something we used to do-pitting Internet against traditional networking-and instead take an agnostic view on what applications are best for customers."
This means supporting existing applications, whether frame relay, ATM or IP. "Our philosophy is that there is not one right single tool for the job," Brewer says. "If a customer believes that the Internet is secure and ready, then we can provide them with our Internet backbone. If they want to have a little bit on both sides-an IP-enabled frame relay network, for instance-we can handle that as well."
Overall, WorldCom's network is substantially built out, according to Brewer, who says the carrier has enough bandwidth, enough fiber, enough equipment to support its transmission capabilities. Where the company wants to continue to make investments is on the edge. WorldCom recently acquired assets of bankrupt Rhythms NetConnections, with which it plans to offer DSL to businesses as a lower cost T1 alternative.
Also, WorldCom is just now rolling out commercial Gigabit Ethernet private line service, though Brewer says the company is not as convinced as some of its competitors about the likely demand for it. And with the carrier's MMDS (multipoint multichannel distribution service) spectrum, WorldCom says it continues to test services, though it has yet to offer fully commercial products.
Overseas, WorldCom continues to do the more traditional type of network construction, which Brewer says is basically complete in Europe, where the carrier reaches all major metropolitan areas. In Asia-Pacific, though, construction is underway to reach new cities. In new markets yet to open up, Brewer says the carrier will offer only data and IP services.
While WorldCom may want to accommodate existing customers wedded to more traditional applications, the carrier's focus on IP in this case underscores WorldCom's view of tomorrow's network. "The growth opportunity is in data and valued-added services," Brewer says.
-Ted McKenna
--------------------------------------------------------------------------------
Walter McGrath
General Manager
BELD
If there are two words to describe technology deployment plans for BELD (Braintree Electric Light Department), they are tried and true. BELD provides electricity, Internet access and cable TV services to the town of Braintree, Mass., an 11,000-home community located about 15 miles south of Boston. The utility began installing HFC in 1998 as part of an automated meter-reading system, and around the time the project was about 70-percent complete, broadband access hit its stride. "No one else was offering broadband access in town and we felt we could do it, since we had the HFC installed already," said Walter McGrath, BELD's general manager. About a year later, after a town vote and a $3.5 million authorization for Motorola headend, set-top boxes and antenna equipment, the project was expanded to include digital cable TV services.
The take rate for both has been strong: BELD boasts nearly 3000 cable modem subscribers and 3650 cable TV subscribers; the cable subscribers were taken directly from AT&T. "We're considerably cheaper and AT&T just started offering digital TV this summer," McGrath said. "They may have just started offering broadband Internet access as well. We're not feeling any effects from that if they are." BELD also has seen success in bundling services-the provider takes $5 off a month if subscribers receive both Internet access and cable services.
BELD does not plan to expand its service area-it owns the poles in Braintree but not elsewhere-but expanded services are on the horizon. "As soon as video on demand is perfected, we'll look at that," McGrath said. "No one's calling to demand it just yet and I'm not sure what, if any, additional infrastructure will be needed, but it's definitely something we're considering." On the Internet side, BELD also offers services such as Web hosting on a limited basis. "We're a small shop, so we don't want to get in over our heads," he said. "We've had some hiccups on the equipment side, so our main goal right now is to add more reliability."
What about basic phone service? "When we went into the cable business, our consultant said we would have to offer all three services-voice, video, data-to be truly successful," McGrath said. "We want to make sure we have everything running smoothly before we look at something like that. Besides, the 911 requirements really scare me."
-Sue O'Keefe
--------------------------------------------------------------------------------
Dennis Hinkel
Senior Vice President, Network & Operations
Cincinnati Bell
As an incumbent phone company serving business and residential customers in Ohio, Kentucky and Indiana, Cincinnati Bell focuses on delivering good, basic telecom service. So it naturally tends to take a practical approach toward newer networking technologies.
Through its ZoomTown Internet access division, Cincinnati Bell now offers ADSL service in 82 percent of its service area and has signed up 55,000 subscribers. How much more extensively it plans to offer DSL depends on demand within individual neighborhoods, according to the telco's senior vice president of network and operations, Dennis Hinkel, who says his company works with partners to offer new services with DSL.
Video, for example, is offered through a partnership with Intertainer, which provides videos on demand. Other applications offered by ZoomTown include connection to business LANs from the home and online training. To permit the delivery of higher bandwidth applications, Hinkel says the company is studying VDSL (very high data rate DSL), although it hesitates to implement it because of high capital costs.
Deploying VDSL would require all new DSLAMs and more of them to reach the same number of subscribers, since VDSL technology cannot run as far along copper lines as other types of DSL. So while the 26 Mbps transmission from VDSL would allow the delivery of more video channels over the same stream, Cincinnati Bell believes partnerships with companies such as Intertainer open up possibilities to deliver transmissions more efficiently over networks.
"We think that products will continue to evolve, and these paths will start to come together, so maybe you won't have to deliver 26 Mbps, and if you do you can offer that many more services across it," Hinkel says. "Services like Intertainer could deliver a video stream that is more comparable to what you get on embedded cable networks. So we think there's promise on both ends."
Whether by offering video or some other application, the carrier sees greater choice as the way to generate more revenue from its DSL service, as well as sign up more customers. "Obviously services are the key to unlocking the next wave of revenue growth," Hinkel says. "We have effectively used bundling to get customers to sign up for customer calling type services, high-speed Internet access on top of voice services, and more."
-Ted McKenna
--------------------------------------------------------------------------------
Fred Harris
Vice President of Design
Sprint
Ask Fred Harris how Sprint will continue to improve its network and offer advanced services amidst declining capex, and you'd better be prepared for a math lesson. Take, for example, his challenge to vendors regarding Sprint's plans to migrate to an optical mesh core. "I'm setting an interesting paradigm for vendors: If I am going to spend $10 million with them over the next year, I want them to identify where I can save $10 million in capital and operational expenditures," Harris says. "At least two of them have been able to show me how they can do that over a couple of years, and I've asked them to break that down to an annual recurring strategy. So essentially it doesn't cost me anything to make that change, and I have the ability to reduce the capital infrastructure in the network core."
Harris isn't shy about Sprint not being first to the table in breakthrough technologies like optical switches, which Sprint has not yet deployed but likely will in 2002. "I think we had some concerns about the service-assurance aspects of all-optical networks, but the fact that there have been some all-optical solutions deployed is very encouraging," Harris says. "In the optical layer we feel we have the luxury of being able to wait and see how those technologies emerge."
One area Sprint isn't holding back on is softswitch technology. Early last month, Sprint inked a $1.1 billion deal with Nortel Networks to replace Class 5 end-office switches with Succession softswitches and MG media gateways-the largest VoIP deal to date. Although softswitches have been criticized because of scalability issues, Sprint feels it has enough experience with the technology to work out any potential kinks. The carrier originally used Telcordia softswitches in its ION project, which was cancelled in October.
"We are going to push the envelope with softswitch technology," he says. "They will handle all of the voice growth in the Sprint network from now on. We know how to make the technology work, operate and maintain it, and provide service assurance. Some people will criticize this, but the proof in the pudding will be that we will reduce our opex spending significantly in the voice category."
Sprint's goal is to have its voice services with the new softswitches running over an IP backbone beginning in late 2002. "That means we're going to have to provide traditional QoS over that environment. Based on ION, we think we know how to do that," Harris says. Look for value-adds from Sprint to include storage-enabled networks (which the company is also pursuing for internal use as well as a service) and telecom portals, which Sprint is developing internally and hopes to have available by the end of 2002. And don't forget anything over IP-IP VPNs, IP Centrex capabilities, voice over the Internet. "Then there is a host of potential value-added services that can be provided through policy enforcement and a variety of systems behind those that will allow you to tailor services specific to customers," Harris says.
And what about services such as voice over DSL? "We have interest in it, but it's not high on our list," Harris says.
-Sue O'Keefe
--------------------------------------------------------------------------------
Scott Clavenna
President
PointEast Research
Q. Is there one new technology that you feel can make a stellar impact on a service provider's cost/capacity/performance issues?
A. The one technology that has done it and will continue to do it is optics. There is nothing new to search for; nothing can touch optics for the foreseeable future. We're in the process of evolving optical technology to add more networking features. We've gone from using optical fiber to replace copper, to using DWDM to replace fibers, and now we will be focused on networking those wavelengths and extending their reach. That requires lots of new technological advancements in amplification, wavelength switching and tuning, and wavelength management and monitoring. That's what is bringing the focus back on components companies, because they will initially drive these advances.
Q. What does the so-called bandwidth glut mean for investments in long-haul and core equipment? Which vendor is in the best shape and why?
A. Carriers are looking to optimize their infrastructure investments in the coming years, so products that exploit what is already deployed will succeed. Optical switching systems that both switch wavelengths and groom electronic signals within wavelengths have the strongest market opportunity over the next five years. All-optical switches in the core have two strikes against them: They are unproven technically and carriers still do not have operational methods in place to manage all-optical nodes with the same kind of monitoring and test access that they have with electronic solutions. The vendor with the best position today is clearly Ciena. Ciena's challenge will be to compete effectively with Cisco as that vendor improves its service provider reputation and adds quality DWDM transport solutions to its product suite.
Q. How will carriers address the bandwidth dearth in the access network?
A. Access is paramount to carrier success: Consumers have the ability to drive new services and bandwidth in ways that will ultimately dwarf the business/metro market. The consumer drove the PC and Internet revolutions and can drive a whole new wave of bandwidth demand with broadband. What's needed, therefore, is very low-cost, scalable access platforms. I like some I've seen for cable TV networks, and a couple of vendors are developing new digital loop carriers, turning them into "optical loop carriers" if you will.
-Sue O'Keefe
--------------------------------------------------------------------------------
Steve Craddock
Senior VP, New Media
Comcast Cable
Q. Where is Comcast in its DOCSIS 1.1 and PacketCable rollout?
A. We need DOCSIS 1.1 to ensure QoS for real-time voice, gaming, open access and tiered services, but it's not required for best-effort surfing. DOCSIS 1.1 is not only a requirement in the residential arena, but an absolute necessity in the business segment where you are trying to displace T1 and 56K lines. Between DOCSIS 1.1 and DOCSIS 2.0, you can prove you're carrier class. Right now, we are planning for 1.1 service. I have a RFI for DOCSIS 1.1-enabled voice. My goal is to evaluate PacketCable-compliant softswitches for eventual lab and field trials. We have budgeted money to install DOCSIS 1.1 CMTSs (cable modem termination systems) next year. Those subscribers who want new services will be placed onto the new system and the basic Web service will be kept on DOCSIS 1.0.
I have been baking off next-gen CMTSs in the lab for the past few months. Not only do they have to be 1.1 and PacketCable compliant, but have a next-gen engineering mindset with a midplane design, hot swappable RF connections with 1XN seamless failover. We are looking at CMTSs from ADC, Cadant, Cisco and RiverDelta/Motorola, and the usual suspects-Terayon, Riverstone and Tellabs. Although we have not released results yet, some of these systems are really carrier-class.
Q. Where is Comcast in its VoIP migration?
A. We gained some Class 5 switches through a few acquisitions. We have a 5ESS via Media One/AT&T up in Detroit, which was already a full switch for about 100,000 lines, so it makes sense for us to use that capacity. They were going to serve that with conventional HFC lines, but it's pretty expensive. One of PacketCable's profiles has a GR.303 implementation that creates an IP access line from the home to the headend, and a GR.303 gateway to turn IP into circuit-switched traffic, which is terminated on the regular line connection. If everything goes right, the IP part is almost invisible. To nail up dedicated spectrum in an expensive environment, you treat the voice call as a data call with IP packets. At the CO and headend, those packets are changed back to circuit switched.
Q. Are softswitches ready for MSO prime time?
A. We ran two trials for softswitches-one with Lucent's PathStar before they killed it and an ongoing trial with Telcordia and Cisco in Philadelphia. We have an RFI for IP voice softswitches, but they have to be PacketCable-compliant. Many first-gen softswitches were targeted at CLECs and international markets, but with the implosion of that market, these guys decided the MSOs are the new opportunity. Because there's more than a trivial difference between features of a Class 4 and a Class 5 switch, a lot of these guys are having trouble becoming Class 5 and PacketCable compliant. This leaves room for guys like CedarPoint or Israel-based Gallery IP that are PacketCable compliant. Gallery IP's story is kind of amusing in that they built a switch, went to CableLabs and said, "Will you test it?" Sure enough, it was PacketCable-compliant. CedarPoint integrates feature and announcement servers right in the blade, which is not only scalable and flexible but also is great in terms of security and efficiency, because it has an optical backplane. While the jury is still out, integrating it in the blade and one box is promising. Since they are going directly after the MSOs with a PacketCable-compliant system, they have cut six months off their production time. Syndeo is another one I like.
Q. What is your take on CableLabs' DOCSIS 2.0 specs?
A. When we examined both Advanced PHY layer proposals (i.e. FA-TDMA (frequency agile-TDMA) and S-CDMA (Synchronous-CDMA)), we asked Terayon and Broadcom to collaborate on one spec that utilizes both schemes. Out of that came an Advanced PHY layer that incorporates both schemes that we call DOCSIS 2.0. I am very excited because it gives us the robustness and throughput for video telephony and symmetric business services. We have given vendors a period where we will entertain partial qualifications that have one or the other scheme. We will put both in because you can turn one on or off. Broadcom's FA-TDMA has some scheduling algorithms that are really cool, but there are many places where it won't work because it's based on FEC (forward error correction). With short packets and impulse noise, S-CDMA is better suited in that environment. Each scheme will give the MSO the best performance in a certain set of circumstances.
-Sean Buckley
--------------------------------------------------------------------------------
Jack Biery
Vice President Engineering and Construction
AFN Communications
As a carrier's carrier, AFN Communications has lit about one-fourth of a super-regional, 8000-route-mile network that will have a triangular footprint extending from Albany, N.Y., across to Chicago and down to Greensboro, N.C. AFN's target market is providing private line (DS3 through OC-192) and protected/unprotected wavelength services in secondary and tertiary markets. "DS3s are a solid seller. It's sort of a unit of exchange right now. But we're also seeing a lot of traction for OC-48 service and believe that the market for 2.5-Gbps wavelengths will grow for data traffic," says Jack Biery, AFN's vice president of engineering and construction.
With an average circuit running between 200 and 300 miles, the ability to add/drop channels flexibly and economically is particularly important. Moreover, AFN believes it's harder for a carrier to attain efficiencies and reach profitability price points if, for example, it must buy one system for 2.5-Gbps optical transport, another system for 10-Gbps transport, and perhaps a third system if it needs ultralong haul capability. Channel planning or design tools also would help carriers. "It would be nice to say these are the cities we want to light up, here's our route and fiber type, and we want to drop this much traffic in these cities. Now give me a channel plan model to do this," Biery says.
AFN was launched when a handful of energy and telecom companies contributed certain assets such as fiber (some of it lit using Nortel and Cisco/Cerent SONET equipment) and some customers. AFN is in the process of transforming that infrastructure by buying a variety of optical equipment from Ciena, specifically the CoreStream DWDM product, MetroDirector K2 SONET-based multiservice switch (formerly the Cyras platform), and the CoreDirector intelligent optical switch. The carrier recently turned up its first route using CoreStream and is in the process of deploying CoreDirector. AFN is currently generating revenue and claims to have sufficient funding to take it to a cash flow positive stage.
-Sam Masud
--------------------------------------------------------------------------------
Dan Sheinbein
VP, Network Architecture & Development
AT&T
Q. At what technologies are you looking closely?
A. We have lots of new fiber going in and before the year is out we're going to put in intelligent optical switches that provide point-and-click capability for provisioning services. We'll also install MSPPs (multiservice provisioning platforms) in the access (including customer premises) and metro core-devices that take lower speed inputs and multiplex them up to the optical layer. These MSPPs will have the same point-and-click capability that's in intelligent optical switches. Because MSPPs are deployed closer to customers, it's very useful if you can point and click, and provision services on an end-to-end basis. This is how we will be able to achieve significant operational cost savings in addition to capital reductions.
Q. What's important as far as DWDM systems go?
A. In the DWDM area, we're interested in more wavelengths and in ultralong haul systems. And it doesn't matter if this is provided by systems that work in the C, L or S bands because the issue is more wavelengths over longer distances. We began by putting in 32, 64 and 80 wavelengths, and by year-end we'll introduce 160 wavelengths. More wavelengths and capacity are good, but the important stuff is in the intelligent optical switches and MSPPs.
Q. Do other technologies interest you?
A. We are very interested in IP technologies such as IP VPNs with MPLS and QoS. Also, and this goes back to what I said about MSPPs, we want to support Gigabit Ethernet/10 GigE in the metro as well as in the WAN. We want to support native Ethernet transport as well as perform GigE switching. Today, we support GigE buried within SONET. We are looking for GigE switching capability that we hope will be even cheaper.
Q. Are you interested in softswitches?
A. Very interested. We have some in our lab right now and are looking at them for both Class 4 and Class 5 functions. These functions require very different capabilities. In the voice world, we know that the circuit-switched network will be with us for a while. If in the future even half the locations have IP capability and half have circuit switching, there will still be the need for calls to traverse between these networks. Because of this hybrid environment, softswitches and media gateways will be a major enabling technology. The issues that need to be addressed are reliability and scalability. Reliability means providing the same services that you currently provide with circuit switches, and it's not just features like call waiting but things like 911 service. We of course have a very large network that handles lots of calls, so these softswitches and gateways have to scale.
Q. Of the technologies you've mentioned, do two or three have priority?
A. Many of them are important. Intelligent optical switches, MSPPs, softswitches, MPLS-based IP VPNs are the hot buttons. We're also focusing on IVR (interactive voice response) systems as a major technology area.
Q. What are some of AT&T's important new services?
A. Obviously IP is the industry's growth direction. At the same time, frame relay and ATM services are growing between 50 percent to 100 percent; third, there's GigE. Obviously, we will continue to support our critical voice services.
-Sam Masud
--------------------------------------------------------------------------------
Chris Nicoll
Vice President
Telecom Infrastructure
Current Analysis
Q. What does the so-called bandwidth glut mean for investments in long-haul equipment?
A. There is a glut of potential capacity in the long-haul network, not an actual bandwidth glut. BIG difference between open slots in a DWDM system and lit wavelengths that are not being used (or sold). Some networks, such as Sprint's, really are out of optical capacity and need to be rebuilt/expanded. However most networks have fiber that could be lit, but is not for a couple of reasons: Providers don't need the capacity right now or it's the "wrong" type of fiber.
What we have not seen yet, and I thought we would have by now, is a clear advantage in time-to-revenue for service providers that are building out intelligently switched optical networks. Broadwing can very inexpensively provision wavelengths across its architecture, due to the nature of Corvis equipment. However, other vendors can do the same with older systems, they just have to throw a lot of manpower and equipment at the problem. When it gets to the point where that manpower and equipment is too expensive, then I think we will start seeing more of a move to the new optically switched architectures.
Q. Is there any one new technology anywhere in the network that you feel has the ability to make a stellar impact on a service provider's cost/capacity/performance issues?
A. In the very near term, there are two areas where service providers are starting to make investments with dramatic bottom line results: their SONET and ATM networks. Both of these traditional service areas are benefiting from next-gen hardware architectures that reduce the size, power and cost of SONET and ATM switching systems. Companies such as Gotham on the ATM side and Ocular on the SONET side are showing service providers that it is possible to drop these new systems into existing networks, nearly transparently, and continue to offer the exact same services at one-tenth the power, one-half the space and one-fifth the cost.
From a future perspective, intelligent photonic switching holds the promise of creating end-to-end wavelengths when and where needed. Service providers have enormous flexibility in routing these wavelengths in this type of network, and the time to revenue (not just time to market for a new service, but how quickly revenue can be gained from the service itself) is dramatically reduced.
Q. How do you feel about multiservice boxes?
A. With the current economic climate, carriers are not putting systems in on the "promise" of future capabilities or services. Vendors have to show that they are solving a specific problem the service providers have, and most of the god boxes are just not best in class at solving any one particular problem. The metro market is littered with god box platforms that have not yet lived up to their early promises: For example, Cyras' K2 was having difficulty establishing traction in the marketplace on its own. The K2 will be successful with Ciena's ownership and guidance, but I expect we'll see less true multiservice functionality and more optical switching and grooming to complement Ciena's other products. This is what Sycamore did with the Sirocco products, focusing on the complementary switching and grooming functionality which strengthens Sycamore's overall solution.
-Sue O'Keefe |
