Management Briefing

“The days dwindle down to a precious few…” goes the song, talking about autumn.  We’re sailing into the last autumn of the millennium, and it’s logical to take this time just before that last autumn starts to consider what might be developing in the months to come.

Regulations are a good place to start, and since the regulatory picture is probably the largest factor influencing the current networking market, it’s a smart place for us to start as well.

The FCC is likely, sometime this fall, to publish the final rules that would govern the way that RBOC subsidiaries would be allowed to deploy advanced services like IP and xDSL.  Our readers know that the FCC had previously indicated that the RBOCs would be able to dodge wholesaling advanced infrastructure only if they spun it out into a separate subsidiary.

The specific way the rules come down with regard to how that subsidiary will operate will determine the nature of xDSL competition and a lot of other things about future networking.  A few key issues must be decided:

1.       Will an RBOC CLEC subsidiary be allowed to deploy data-only xDSL services even if the state PUCs in some states have decreed that xDSL is a regulated service that must be wholesaled?  This would have special impact on SBC, since its largest market (California) has such a rule.

2.       Will an RBOC’s CLEC subsidiary be able to use xDSL to transport voice services provided by the ILEC portion, or would even such transport carriage be considered “provisioning” voice under the Telecom Act?

3.       Would data services and equipment already deployed by the RBOC’s ILEC be transferable to the CLEC subsidiary, or would this be considered a transfer of assets to evade wholesale requirements?  Conversely, would an RBOC who had deployed some technology like xDSL or ATM in a subsidiary be able to pull it back into the ILEC?

4.       What would be the relationship between an RBOC’s in-region advanced services CLEC and any out-of-region operation it might undertake?

5.       Will the FCC assert its jurisdiction on all of the issues related to data services and the division of an RBOC into an ILEC/CLEC?

These questions would not only impact the competitive landscape directly, they’d also impact the question of whether the RBOCs would form a retail CLEC subsidiary, as some have proposed.  That would clearly have a huge impact on the industry.

The idea of an RBOC split-up originated with some competitors who believed that unless the RBOC’s ILEC became what would be in essence a wholesale-only entity, with retail operations separated as a CLEC, the RBOC would never really assure open wholesale access to infrastructure without long delays in Federal courts and other regulatory venues.  The theory was that such a split would instantly resolve the compliance problems associated with the Act, because the RBOCs couldn’t sell services to end users without compliance.

Lately, the concept has gained favor in the RBOC community, where it was felt that the split would demonstrate that the current TELRIC pricing and Universal Service subsidies would not cover real costs.  That has long been argued by the RBOCs, who feel that their retail profits are being used to subsidize competition because wholesale rates are set too low.  Removing the retail operations to a hands-off subsidiary would make that subsidy impossible.

Recent developments in Pennsylvania may be driving things to a head in this area.  The PUC there has ordered a price reduction of more than 15% on UNEs, which in itself would probably offer Bell Atlantic a powerful incentive to break into a wholesale/retail pairing of companies.  But leaving nothing to chance, the PUC has ordered them to do that within the state.  Having this level of separation would almost automatically insure that Bell Atlantic would meet the requirements of the Telecom Act to enter the regional toll market.  The rest of the RBOCs, still fettered, would probably act quickly to follow Bell Atlantic’s lead, and if breaking up seems a path to the FCC’s heart with respect to compliance with the Act, so be it.

Still, it’s issues like the Pennsylvania decision that make the fifth question above so pressing.  Can Bell Atlantic reasonably break up in one state alone?  Will piecemeal regulation of specific advanced services issues stymie the whole xDSL and IP services adventure?  Certainly uncertainty there would push the RBOCs back to voice services competition, but that might have the same effect in the long run—promoting separatism and compliance with the act as a strategy to control wholesale costs and retail profits, respectively.

One way or the other, though, it seems very likely that the RBOCs will end up unfettered from the regional toll restrictions of the Act by early next year, and that they will be preparing for the competition this fall.  It also seems likely that they will be splitting off CLEC subsidiaries, though the exact relationship between those subsidiaries and the native ILEC is yet to be determined.

In the equipment space, this is a key issue for the DSLAM vendors.  Look at the options here:

1.       If the RBOCs were to decide to deploy xDSL in a subsidiary, the voice-over-DSL play is the clear winner in the marketplace, and the concept of integrating data features into DSLAMs is also a winner.

2.       If the RBOC were to decide to deploy xDSL out of the ILEC, then everything depends on whether the RBOC is willing to wholesale data.  If not, voice-over-DSL is still a winner and data in the DSLAM isn’t.  If the RBOC does wholesale data, then it’s not clear whether voice-over-DSL wins, but data in the DSLAM does.

The next issue for the fall is the final death of the “convergence” market driver.  What will kill it off are changes in the new-generation competitive carrier space.

This fall, we expect to see the CLEC market start to shake out.  CLECs with no facility value-add will clearly not survive no matter what regulatory trends emerge, but the formation of an RBOC CLEC subsidiary will be quick death for them.  But even the so-called “data CLECs” will have problems.  The challenge is the return on capital that simple access or transport services offer.  Bit pipes can’t be feature differentiated, so their purveyors will have to duke it out on price.  That means the margins for simple access via xDSL, and simple transport (long-distance or long-haul) via anything will begin to commoditize.

The current IXCs are starting to hike their rates for raw digital bandwidth, knowing that their current T3 and higher offerings are being consumed by competitors over users by at least a 4:1 margin.  Moving the users to ATM (via the new ATM integrated service offerings) will move real users to a form of bandwidth where the service provider is already taking advantage of statistical gain.  The competitors, who need raw bandwidth so they can leverage that statistical gain to raise their own margins, will be left behind.

That will move the ISPs away from the current IXCs and more to new-gen carriers, but it will also insure that the margins those carriers can expect on bandwidth will be minimal.  Thus, the new-gen players like Level 3 will have to get more into the local access market, and stress services more than just simple leased lines or telephony.

IXCs this fall will be moving to promote their own shift to higher-level services.  AT&T has several beta customers for its new MPLS-based VPN service, and at least one will be going public this fall.  The publicity on the service could force the other IXCs to stop diddling around on facility-based VPNs and get on the bandwagon.  Cisco is certainly pushing them to move, because it has a lead in that space as the provider for the AT&T network.  Lucent, still digesting Ascend, will have to move quickly to keep control of the key RBOC accounts.

Both these trends are going to push voice service enhancement out of the “IP convergence” arena into the feature differentiation arena.  That doesn’t necessarily mean that IP won’t be a play at all in voice, but rather that the value of VoIP will be not as a better transport option for voice but as the basis for a better feature platform.  If most really good voice-feature-driven new-gen voice (NGV) players offer a variety of transport options (which we think they will), then VoIP is a side show.

Our subscribers know already that we believe the feature drive in NGV will be based on XML definition of cooperative service relationships and feature logic, with Java playing a role in defining specific behaviors of devices.  There will be a number of announcements from new startups in this area during September and October.  A major vendor in the voice space will also be adopting a flexible feature architecture in this same time period.

NGV players will probably figure in some high-profile acquisitions this fall, too.  Alcatel and Ericsson are both un-invested in the NGV space, and neither can really afford to have a hole in their product line there.  Of special interest is Alcatel, whose aspirations in the xDSL space put it in the forefront of digitizing key customers, and thus creating the market segment to which NGV technology would most likely be targeted.  Nortel, as well, is quiet in the NGV area so far and rumored to be courting a new player to acquire.  Siemens/Unisphere has Castle, but has been all too quiet in the months since its formation, and must now try to win back lost momentum.  Lucent’s SoftSwitch approach has gathered some momentum with the Level 3 deal, but still has to demonstrate a more sophisticated strategy for feature creation than simply using custom programming and APIs.

The IXC play on VPNs will probably begin to stabilize the “facility VPN” equipment market this fall, as well.  Tunneling VPNs are, in truth, simply alternatives to private lines created by playing on the zero-usage-cost nature of Internet connections.  They don’t populize IP networking.  Facility VPNs, which act as true partitioned private IP networks created on public facilities, would accelerate the current network outsourcing trend and also increase the data market overall.

The VPN space is shaping up to be a Cisco/Lucent war, but neither player has been showing all its cards.  Cisco has been reluctant to accelerate private networking’s decline when they have most of the revenue benefits.  Their deal with IBM should make them a bit more secure here, as we’ll cover later in this issue.  Lucent, as previously noted, has been preoccupied with restructuring in the wake of the Ascend acquisition, and is at risk to lose momentum if they don’t pick up quickly.

This is a logical time for new players to emerge.  Newbridge, having acquired Northchurch and having a stake in Vancouver startup and VPN access player Abatis, may be planning to take advantage of the temporary confusion among the key players.

There’s also room for other startups to emerge.  One player, FirstStar, has announced that it will be using the new Intel network processor chip in a multi-product VPN architecture.  The availability of these enhanced network chips from C-Port, IBM, Intel, and Maker (to name a few) is likely to trigger additional plays in the VPN access area, since this type of product is ideal for a VPN edge solution framework.

Things are going to be getting confusing, but there may be hope that some of the confusion will be dispelled by a new attitude in the media.  The boom period of networking publications is passing quickly, and readers’ dissatisfaction with the usefulness of the press is growing.  Some major mags are likely to drop this fall, and those that don’t die may become somewhat more interested in providing value.  But hype dies hard, gang, and even threats of death can’t make a vapid process insightful overnight.  We’ll have to watch and see how the survivors do at learning a hard lesson.

For the users, the fall will bring the final realization that private networking may be dying.  Some organizations will try to create a second wave of investment with things like packet voice to load voice traffic back onto their networks, but most companies are already finding that the declines in phone rates are cutting costs more rapidly than any capital network commitment could hope to do.  Inevitably, private networking will slip into decline.  We expect to see enterprise sales of WAN networking equipment starting to plateau within three years.

Service provider equipment may or may not keep up, at least that equipment in the service provider data space.  A public data network is more capital-efficient than a private one, which is one of the arguments for creating public networks in the first place. 

Interesting times are in store for the millennium, and the Y2K problem has nothing to do with it!

In the Know

The concept of VPNs is perhaps the single worst-covered thing in all of networking.  The problem lies in the fact that the weekly trade press has a readership that’s predominantly end-user in nature, and in fact not even network professionals.  Such a group is likely to be Internet-infatuated, and interested in private lines and similar things—what they’re already used to.  The idea that a public IP service would be based on facilities technology and offer segregated private network emulation is hard to deal with in that forum.  As a result, the media has focused on the kind of VPN that a casual reader might like—the kind you create by simply building Internet tunnels from site to site.

Real VPNs aren’t this at all, as virtually every professional network planner knows.  But what is really involved in a “real” VPN isn’t as well known as it should be, largely due to the ineffective coverage of the topic.  Since we’re going to start seeing “real” VPN wins this fall, it’s time to explore the issues in more detail.

Sorry, Internet readers, but this article and its successors on this topic are for subscribers only.

Strategies

There is going to be a new public network; we all know that.  What we don’t necessarily agree on is how that new network will differ from our current one.  We also rarely agree on how soon that network will come about.

In truth, the issues are related.  Networking is an economic engine, and changing it costs money.  In the case of the new public network, the money would be in the form of investment in new infrastructure, and the source of the money would be revenues from services.

A simple view of the future could be stated in those terms.  Old stuff ages at a given rate (its depreciation rate).  As it ages, it’s replaced by new stuff.  If technology improvements allow new stuff to be more efficient by adopting a different paradigm than the old, the paradigm of networking will shift at the rate that the new stuff can be adopted.  This all works even if the services the network offers don’t change a whit along the way.

What new services do is introduce into the picture a little more pressure to evolve.  If new equipment is needed to support new demand, the pace of network change will move faster than depreciation along can justify.  Users will buy new gear to support the increased revenue opportunity.  In addition, the percentage of depreciated-out stuff that gets replaced by new technology instead of just updated old technology will increase.  The result will be a quicker transition to what we might call the age of the future.

That’s the dilemma we face in networking today.  We have an equipment plant in the public network worth about a third of a trillion dollars, and it’s depreciating away slowly because the gear has a 15- to 20-year cycle.  The rate of change that depreciation alone brings about would keep us from changing out even half the existing infrastructure by the end of the next decade—2010.  We have a hope of new demand in the data space, but it’s hard to say how fast revenues there will rise.

But isn’t the traffic on the Internet doubling every 100 days?  Maybe it is, but the dollars aren’t, and bits don’t buy bits, bucks do (as they say).  Since the ISPs are pretty uniformly unprofitable, we’d have to look elsewhere for rapid revenue growth in a new demand area.  Frankly, that’s a tough order, since nothing seems to be exploding out of the horizon.

Data is the hope, but it’s a deferred hope based on current spending patterns.  In fact, the fastest that we can hope for a revolution in data revenue is the second half of the next decade.  That lets us divide the future into some phases, and it’s that phase structure that we’ll use here to talk about how vendors may be facing the changes and opportunities of the new public network.

Voice and Data, New and Old

At the likely pace of data growth, we’ll hit parity between old-generation infrastructure and new-generation infrastructure by about 2007.  By about 2005, future data revenue growth will be promising enough to have insured that most depreciated gear is replaced by something more data-ready.  Thus, we can set the voice/data dividing line conveniently in 2005.

One thing this immediately suggests is that there are a lot of players in the market who have seriously anticipated demand.  The primary source of new data revenue in the public carrier space for the next five years will be the translation of current enterprise private network applications to an outsourced public network model.  Frame relay, ATM, and IP VPNs will all benefit from this shift in direction, though modestly.  The result, nevertheless, won’t be a major profit source to carriers because leased-line revenues will fall to the extent these new revenue sources rise.

If data commitments will be growing more traditionally in the next five years, that makes a radical shift of equipment vendor dominance based on data investment unlikely.  Cisco, who has pinned its hopes on public IP, is the one most likely to be disappointed by the pace of data growth, and the one most likely to have to take steps to accommodate it.

The Cisco deal with IBM to pick up the latter’s networking business is an example of one smart thing for Cisco to do.  Enterprise networking will plateau, but IBM’s installed base represents the last bastion of private networking, and a market that Cisco hasn’t completely cemented.  Conversion of SNA networks to IP would create several billion dollars of new revenue for Cisco, and that would help to bridge the gap between the downturn in enterprise networking overall, and the ramp-up of public IP data services.

The risk Cisco has here is that a lot of SNA users will migrate to public services rather than to private IP.  Cisco’s attention to the private network voice over IP space is partly aimed at revalidating the private network, but as we’ve indicated earlier, this wasn’t as bright an idea as buying IBM’s networking business.  The KPMG deal was also probably directed at greasing the skids of uncertain user planners in making a private IP commitment, but we’re not sure that will help much either.  Cisco needs to try to jump-start its service provider business.

One way of doing that was the Cerant acquisition.  Cerant is a fiber-edge player that can be used to penetrate the highly vulnerable SONET infrastructure market.  Because Cerant can do SONET, it can be used instead of more limited old-generation SONET devices as the core element in traditional TDM networks.  Because it can also do ATM and IP multiplexing, it can be used to help lead those same providers into the world of advanced networks.  That’s important for Cisco because they’d like to sell the infrastructure that would surround the optical core and provision those advanced networks.  Cerant could give them a place in the core, and Monterey (a DWDM networking vendor somewhat competitive with Sycamore) would help create a more efficient optical infrastructure than Cerant could do alone.

The Cisco move at the core is vital for Cisco because they’ve already lost the voice space.  Cisco has refused to admit to the continued success of both PBXs and Class 4/5 switches, preferring to rant about the need to replace this stuff with routers.  Since that isn’t going to happen any time soon (at least not in the next five years), Cisco gave Nortel and Lucent a free spin in the game of service provider marketing, with the largest prize on the table being the thing up for grabs.

But there are issues that could advance the deployment of NGV hardware, even though that hardware might not be IP-oriented.  Lucent and Nortel need to be sure that they don’t lose the NGV market, if that market develops credibly.

The driver in NGV is custom calling features.  If data can’t be made to differentiate new-generation providers and help them sustain margins in a competitive multi-service market, then only special voice features can do that.  We’ve noted before that custom calling generates more local exchange revenue than all data services combined.  It stands to reason that if significant competition develops in the local exchange space, and if data can’t be made to perform in revenue generation fast enough, custom calling would become the focus of competition.

Custom calling features could generate more than ten billion dollars per year in new revenues, if the features could be created quickly and tuned to match the exact demands of users in both the business and residential space.  Custom calling features to link voice mail and e-mail, fax and e-mail, and land-line and cellular/PCS voice, could also be a means of providing multi-service pull to maximize profits.

So far, major vendors all appear to have ignored this space.  Lucent and Nortel have done nothing in their announced new-generation products to facilitate feature creation.  Both appear to be relying on conventional switch APIs and custom coding.  That doesn’t advance feature development timetables, or provide any new form of customization, versus standard Class 5 switches.

What’s really surprising is that Cisco hasn’t caught on to this opportunity.  Their acquisition of Transmedia, which is the least feature-centric of the new-gen voice players, and their focus on Summa Four as a simple programmable switch rather than as a new feature platform, both suggest that Cisco has been banking on all voice feature changes to come about not as a driver in the market but rather as an accidental byproduct of the voice-over-IP evolution they so much want to believe in.  In fairness to Cisco, though, their competitors have largely made the same error.

Voice and data, features and calling, core and service edge.  All of these areas of tension will develop in the next decade as competition in the service provider space heats up, and opportunity for new revenues drives equipment players into a feeding frenzy.  But what will happen, and when?

Market Timeline

OK, we have two scenarios for development of the new-gen network market; the competitive dynamic and the security dynamic.

In the competitive dynamic scenario, the CLECs move quickly to a feature-driven competition with the ILECs in the voice space.  This forces the ILECs to accelerate their own deployment of NGV technology, which depreciates the current equipment base out faster than normal.

Because the CLECs have a feature edge in the near term, they can adopt some IP voice technology without major financial risk inasmuch as the features justify the buyer being a bit gutsy.  The same move would encourage the new-gen IXCs like Level 3 to remain somewhat IP-centric in voice services.

The ILECs would still not be likely to deploy IP voice or IP-based transport technology, but they would probably accelerate their ATM deployment as a result.  This, combined with their NGV approach, would phase out SONET faster because it would increase the amount of non-synchronized, stat-muxed, traffic.

The end result of this would be a chance of reaching parity between old and new infrastructure just about the 2005 milestone.  IP and ATM would share transport duty from that point onward, but probably with ATM dominating because of the ILEC deployment.

The security scenario unfolds differently.  Here, the CLECs don’t move to voice feature differentiation but rather bank on the xDSL “data CLEC” play.  This accelerates RBOC deployment of xDSL in competition and drives down margins on the xDSL services.  There is a modest increase in near-term Internet revenue because of this, but a larger increase in transport costs, so the ISPs don’t gain anything.  But the CLECs have tilted at a major windmill with xDSL, and by the time they realize what’s happened it’s too late for them to get into the voice feature business; they lack capital resources.

The RBOC play on xDSL accelerates ATM deployment by the RBOC but doesn’t touch the voice switch products much at all.  As a result, neither those products nor the SONET network phase out at any rate faster than depreciation, and we end up with a TDM-dominated backbone way into the second half of the next decade, perhaps even all the way to the end.

New-gen IXCs quickly get forced into ATM or dark fiber transport under this model, since there is no real IP play at the edge of the network and no features to justify IP to the buyer with.  IP is relegated to a pure overlay service play, developing only as something you do over an ATM infrastructure.

It’s clear that the security scenario creates a big win for the traditional equipment vendors and a bigger one for the traditional service providers.  Given the fact that we can’t create instant data demand, it’s clear that this scenario will come about unless pressure is created to validate new voice features quickly.

Internet hype has created a monster that, ironically, seems likely to feed more on those who have promoted it the most, than on those they sought to displace.

Down the Line

Next month, we’ll continue our series on the insides of a VPN.  We’ll also recap the Interop issues, and what they might mean in assessing vendor directions for the new year.

NETWATCHER Index Page