Copyright © 2001, CIMI Corporation, Voorhees, NJ, USA. All rights reserved. This issue may be printed or stored in Web format for personal, non-commercial, use, provided that the entire issue including this copyright notice is reproduced and included. Portions of this issue may be quoted, printed, or stored providing that the subject portion is annotated with the issue identification above, and is attributed to the copyrighted material of CIMI Corporation. Other publication, reproduction, electronic storage or retrieval of this material, in whole or in part, without the express written consent of CIMI Corporation, is prohibited.
This section is always reserved for subscribers. Next month, it will feature a review of the service switch market.
Not a good month on
Wall Street, folks. The
optical sector continues to deteriorate (as we predicted) based on both
the fact that it was substantially over-hyped to start off with, and
that access reform is still waiting in the wings.
The problems with the major vendors in telecom/datacom continue
to mount, with Nortel and Marconi both weighing in negatively in stock
performance at the end of the period.
In the regulatory
front, things are still in a state of flux.
The referral of the Tauzin bill to the House Judiciary Committee
apparently didn’t allow the latter to do a re-write of the bill to
incorporate their wishes. Instead,
the Judiciary Committee voted unfavorably on the Tauzin bill and
favorably on an amendment that would have created a compromise—Bells
get data, Bells get more stringent regulation, probably Bells lose voice
for a longer time.
Unfortunately,
while this compromise is likely to be favorably viewed in the House, it
is less popular in the Senate. Hollings,
new head of the Senate Committee responsible for telecom issues, is well
known as an opponent of the Bells. He
held a hearing on the 19th of June that was all but a hate
session against the RBOCs; the witness list made it clear that a
balanced view was not a priority here. Thus,
it is very possible that the bill would get bottled up in the Senate
even if it were to pass the House. On
the other hand, it’s also possible that another compromise could be
negotiated.
The FCC, given
statements by Powell, now seems to realize that the release of the RBOCs
into the interexchange voice market will kill the IXCs.
While that might suggest that he believes that this release
shouldn’t be handled as quickly as the Telecom Act would permit (or as
quickly as he himself suggested it should be brought about in his speech
in January), it’s not clear whether he proposes that the FCC would
drag its feet in issuing new long-distance approvals.
Suppose that the
Tauzin bill dies, and that no other Congressional regulations emerge?
What would then happen in the market?
We can make up a kind of revised “truth tree” based on the
options:
1.
No legislation, and no new
action by the FCC. The
packet unbundling order (99-238) remains in force.
This is a kind of “status-quo” play.
2.
No legislation, no new FCC
actions, and 99-238 is overturned by the courts.
This (obviously) generates the worst-case scenario.
3.
No legislation, but the
FCC acts to reinstate the separate subsidiary exemption granted to the
RBOCs by the FCC but overturned on appeal, and 99-238 stands.
This is the best of the non-legislative possibilities.
4.
No legislation, FCC acts
to reinstate the separate subsidiary exemption, but 99-238 is
overturned. This is an
unlikely case, we think, because any grounds to reinstate the subsidiary
order would seem to validate 99-238 as well, so we’ll leave this out
of further consideration.
The best-case
outcome under case three would give the RBOCs the same regulatory
framework that existed on June 1st, 2000, when Project Pronto
was kicked off. It’s clear
that the conditions at that time were such that the Bells were prepared
to move into access modernization and data services then, because all
were either doing so or preparing to do so.
The question is whether there have been enough market changes to
make a return to that (in retrospect) happy time improbable.
The biggest market
change since mid-2000 was the plummeting in Internet credibility.
Cisco’s stock price is a good measurement of the confidence of
the financial markets in the Internet overall, and it’s clear that by
that metric, the Internet’s not doing well.
If the Internet isn’t the right answer to the question “What
profitable residential broadband services exist?” then we have to
wonder whether there is a
good answer.
We believe that if
the RBOCs are confronted by an FCC environment clearly favorable to data
advancement, but are still facing AT&T-promoted negative lobbying in
the courts and at the state level, they’ll respond with a conservative
build-out plan. Such a plan
would place the highest priority on getting new regional voice and
national data customers as the RBOCs gained entry into the long-distance
market. Business DSL would
also be promoted, but primarily in locations where home-run copper was
feasible, and only in relation to Internet and voice services.
We could expect to see G.SHDSL might do well eventually as a
new-age voice strategy, and it’s possible it might even come to
replace T1. Lucent might do
well here, and Nortel might regret their decision to get out of the
DSLAM business.
The RBOCs would
clearly be jumping on the regional/national multi-service network
business in this scenario, and we would see very rapid ATM switch
deployment in support of this application.
Having a strong virtual tandem/toll capability for ATM networks
would be a key requirement, so it’s likely that the equipment market
for ATM-based voice would do reasonably well.
It might even be enough to save the new-age voice players.
Fiber remotes under
this scenario would be likely to be deployed defensively, meaning that
the RBOCs would want a new-age remote strategy for those green-field
applications involving residential subdivisions, to avoid placing an old
DLC system and then having to replace it if residential broadband were
to be validated in the next five years.
Marconi would probably be favored in this deployment.
If the “status
quo” were to be the course the regulators take, it leaves the RBOCs
with the requirement to share new data infrastructure with competitors,
but provides a packet infrastructure exemption.
Under these conditions, we believe that the RBOCs would take
their cue from the FCC’s pace of letting them into the long-distance
market. If that progressed
at or faster than today’s rate, we believe the RBOCs would let DSL and
data slip back to a holding pattern and instead focus on long-distance
voice.
Fiber remotes under
this scenario would either be defensively deployed as above, or delayed.
With a low-return Internet market the only residential play and
with the continued requirement to share data networks, the RBOCs would
likely be very conservative in what they did to empower users with DSL.
First-cost issues would probably dominate over full-deployment
cost-per-customer in their planning, because it would not be clear just
how far they’d want to go with full deployment.
A regional toll
network based on ATM is still likely in this scenario, but the
network’s overall size and capacity might be more limited, and the
emphasis put on ATM virtual tandem functionality would be greater.
In this scenario, the ATM switch opportunity is more up for
grabs, because the lack of a fast build-out in the DSL space would keep
a vendor like Marconi from eating the whole ATM deal.
Still, with the RBOCs free to deploy unbundled packet
infrastructure, we think the appeal of residential broadband as a future
market would still promote a unified broadband/regional ATM plan, and
the big players would still be favored.
In the final
scenario, the RBOCs lose everything.
Now, it would depend on how aggressive the FCC was in letting
them into the long-distance market.
If the FCC were to slow their long-distance relief under the Act,
the RBOCs would surely go into a longer-term pattern of defensive CapEx
and simply wait for market forces to kill all the competitors—which is
exactly what would happen. Unfortunately,
the regulators would probably get wise to this ugly fate (but too late
to stop it) and create a confused muddle of “save-the-IXCs”
legislation which would muddy the waters for everyone and simply hold
things in suspension longer. This
is our “two-year-industry-sinkhole” scenario.
It seems likely
that any of these scenarios would also see the launching (or
re-launching) of other initiatives designed to stimulate broadband
deployment. The following
have been at least proposed:
1.
Break the RBOCs up into
wholesale/retail entities so “competitors can get a fair chance”.
2.
Pass tax credits that
would provide an incentive to invest in DSL infrastructure despite the
regulatory framework.
3.
Simply order the RBOCs to
provide DSL.
Choice number one
has been kicked around for a number of years, and is the favorite
approach of RBOC arch-enemy AT&T.
The problems with the approach are that it is far too radical to
have any chance of near-term acceptance in Congress, and that it’s
almost certain that it would completely destabilize the system.
A number of studies have reported that the so-called TELRIC
pricing formula applied to wholesaled assets of the ILECs would not
cover cost, and the separation of the Bells into subsidiaries would make
cross-subsidization impossible. In
short, it would help the IXCs and hurt both the consumers and any type
of LEC. Eventually, we’d
have to re-regulate.
The second option,
the tax credit, is essentially forcing the taxpayer to subsidize DSL
deployment. That in itself
might not be a bad approach, providing that the public policy goals of
something like this were spelled out very clearly.
For example, is the goal of DSL the support of today’s
best-efforts Internet? If
not, then it’s clear that the tax credit bill would have to spell out
technology details or the impact might be to foreclose the non-Internet
options. Congress understand
technology? Perish the
thought!
The final approach
has already been tried in Illinois, a state engaged in a mutually
destructive war with its own LEC. In
protest against state regulations, SBC decided not to offer DSL in
Illinois, and the legislature is responding with an order requiring
minimum DSL coverage levels and timetables.
It’s not at all clear whether a state has that authority, in
our view. In fact, it’s
not clear to us that Congress has it.
Of the choices
open, the second seems the most likely to achieve political support.
It has the advantage of appearing to do something while actually
doing not much of anything, a sure winner in the political arena.
As long as this choice was eye candy for the Beltway gang,
nothing much is harmed by the process.
On June 25th, the Wall Street Journal reported that a
bunch of technology CEOs and other bigwigs met with lawmakers to try to
break the logjam of broadband deployment.
One of the measures suggested (surprise, surprise!) was tax
credits.
The problem with
this seemingly acceptable approach is that it doesn’t necessarily
solve the problem (that, after all, should be the first standard to
which any solution is measured). Current
legislation targets tax credits at rural areas, where even with them
it’s not clear that broadband momentum could be developed.
If tax credits were to be applied across the board with broadband
deployment, they’d still demand RBOCs deploy with wholesale/unbundling
obligations on their access networks and on any new service networks
developed to exploit the broadband access.
The RBOCs might see this as barring them from effectively
profiting from the deployment—that it would empower competitors
instead. Thus, they might
still sit down and think things over, while the market slipped further
out of our grasp.
Tax credits are, in
our view, a reasonable way of sweetening the pie if somehow the two FCC
orders of 1999 can be reinstated. If
tax credits are perceived as foreclosing any further action by the FCC
or the courts to restore the subsidiary/packet order dualism of 1999,
we’re all in deep trouble.
We know what a
network architecture is, right? It’s
the design principle that, overall, guides design and deployment of a
network. Do we know what a
service architecture is? Do
we know how it might differ from a network architecture?
Do we know why we might care if it did?
Probably not, so that’s our topic for the month.
Let’s start with
the first of these points. A
“service architecture” is the virtual topology that a service
network presents to its customers. That
service architecture can be mapped to a physical network architecture to
actually build the framework for offering a service.
For example, the Internet is a connectionless service, but the
network or networks that support it obviously rely on physical
connections.
Service
architectures differ from network architectures in that the former
define how the network might appear at the service level, while the
latter defines how it’s actually built.
The Internet example shows how this distinction might be drawn,
but not necessarily why it matters—our third question.
It matters because there will probably exist, for any given
service, an ideal service architecture.
There will also probably exist, for any given service
architecture, a network architecture that is optimal at creating it.
Similarly, for any mixture of services there will be a best-fit
network architecture to accommodate the diverse service architectures
involved, and for any evolving service set there will be a service
architecture evolution that will probably impact the ideal selection of
equipment to support that evolution.
The first question
in service architectures today is the nature of the service itself—is
the service point-to-point, point-to-multipoint?
Is it simplex or duplex? Is
it real-time or deferred? From
questions like this, which we won’t address specifically, comes the
Great Debate; is the service connectionless or connection-oriented.
Connectionless
services are ones that do not create any “stateful” internal network
awareness of a given service relationship.
The packets just flow based on individual routing instructions,
and each conforms to the conditions it finds in each point of the
network it transits. There
is no assurance of QoS, no assurance of delivery.
There’s not even a fixed route, so packets may inadvertently
pass one another, and re-sequencing of packets may be required.
Connection-oriented
services, in contrast, establish a specific resource-to-service
relationship for each service flow in the network.
This relationship, codified by a set of instructions carried in
each node that handles each flow, allows the service to deliver a
specific promise of performance. The
fact that there’s a fixed path through a fixed set of nodes tends to
provide assurance of in-order arrival.
Gee, it seems from
this that connection-oriented services are the right answer, but not
necessarily. Connections
have a price. First, the
stateful relationship requires that nodes in the network be
“activated” with respect to a particular flow before the flow
begins, and deactivated when it ends.
This process of setup and tear-down takes milliseconds when only
one connection is involved, but when tens of thousands are involved,
even milliseconds add up. All
this also takes resources, which are subtracted from the resources
available to handle traffic. Second,
the resources that are committed to a given relationship or flow may not
be available to others even if the target flow/relationship fails to use
them completely. This waste
translates to a potentially higher cost.
OK, maybe we can
state this as a simple trade. Connection-oriented
versus connectionless trades the value of a specific performance
guarantee against efficient use of network resources.
That frames the next question—what applications would tend to
drive that balance in one direction or the other, and what technology
issues would do likewise?
Technology first,
because it’s easy. Connection-oriented
networks are most useful when resources are limited and a guarantee of
performance requires explicit reservation of those resources.
When resources are plentiful, connectionlessness makes sense
because no special procedures should be required to offer a given flow
the service performance it expects.
Lots of bandwidth, or low unit cost of bandwidth, equals
connectionless. Limits on
bandwidth or high cost equals connection-oriented.
Now for
applications. That’s more
complex:
1.
Connectionless service is
favored when information relationships are very short and numerous,
because setting up and tearing down stateful behavior under these
conditions could take far more nodal resources than actually handling
the traffic.
2.
Connectionless service is
favored when information relationships are point-to-multipoint or
simplex, because connection-oriented guarantees of performance normally
require some flow-to-service tuning or control, and that’s hard to
exercise in point-to-multipoint applications.
Self-pacing applications may be an exception here.
3.
Connection-oriented
service is favored when there is no explicit limit to application
traffic (created either by the source/application or by access speed
limits), because applications could otherwise expand their traffic to
the limit of available resources, creating competition for resources and
invalidating the “connectionless-where-bandwidth-is-plentiful” rule
by making sure it’s never plentiful enough.
4.
Connection-oriented
service is favored where part of the flows must pass through a limited
resource and thus must have explicit guarantees made to support that
portion.
Doesn’t sound all
that complex? Well try some
examples.
Voice
communications is a good one. Is
the information relationship short and frequent?
It’s a matter of opinion. There
are point-to-point and multipoint examples, too.
Voice bandwidth use is self-limiting, so we pass there, but voice
service flows on both capacity-constrained and relatively low
cost-per-bit parts of the network. Get
the point?
How about video?
Most video content relationships are probably relatively
long-term. Video
communications (two-way) would probably fit the voice model.
OK, time to make
some approximations, then. Local
exchange voice is probably naturally connection-oriented because of the
combination of local loop congestion issues.
Long-haul voice, in contrast, is probably naturally
connectionless. Video,
because of the higher bandwidth, is probably more connection-oriented
overall. Data in a corporate
sense is connection-oriented, but hypertext Internet is clearly best
served by connectionless networks.
A useful comment
here is that the services that are currently revenue-positive are
primarily connection-oriented for at least a large part of their flow.
That is one of the challenges that face network designers and
planners, of course.
Another challenge
is how the service architecture maps to the network.
Let’s leave aside for the moment the fact that with the
exception of multipoint wireless and some LAN technologies, networks are
connection-oriented at the physical level.
Instead, we’ll consider a more subtle point—the network and
business evolution and its impact on the service-to-network mapping.
A case in point,
and one relevant to current market conditions, is the evolution of the
RBOC network. We would
assert that, had there been no regulatory barriers to impact the
decision and were public IP services immediately credible from a revenue
perspective; the RBOCs would have been smart to deploy IP backbones for
their regional and national networks.
Wait, you say!
This, from an arch-opponent of IP?
We’re not opposed to IP at all, in fact.
We’re opposed to illogic, and the statement just made
represents a kind of “truth-in-a-vacuum” because the conditions that
we cite are not true and cannot be made so. In
the real world, the business framework has a dramatic impact.
In the RBOC case,
look at the facts. The
Telecom Act imposed (probably by accident) a requirement that the RBOCs
wholesale and unbundled everything they sell, in or out of region, voice
or data. This requirement
would tend to discourage the RBOCs from developing infrastructure assets
that, when wholesaled to a competitor, helped that competitor develop a
quick and effective service position.
That’s most easily done when the wholesale asset can be
retailed directly without additional technology elements.
IP infrastructure, available at wholesale rates, would thus
empower competitors.
Then there’s
another dimension, the revenue issue.
Frame relay, transparent LAN, and ATM are all services currently
deployed by RBOCs and based on connection-oriented technology.
These services are creating a significant year-over-year revenue
gain even inside the current set of long-distance restrictions.
In contrast, IP services are not major revenue producers for the
RBOCs at all—most really don’t offer their own Internet.
Thus, it’s a no-brainer that the RBOC would tend to frame the
next phase of network evolution in ATM terms.
Once a carrier
moves in any technology
direction, there’s a significant financial and business/craft practice
inertia that develops. This
tends to keep the carrier on course unless there’s a major regulatory
or financial upheaval which justifies (or dictates) a change.
Thus, the transition from a voice-dominated, TDM-based, network
to whatever comes next will probably start casting itself into stone as
soon as significant investments are made, and the “tactic” for
dealing with current business problems can easily become the
“strategy” of default.
Another factor to
be considered in the service/network mapping is the lack of harmony
within the network itself. We
talk too much about “the network” as though the public network were
homogeneous in business terms. There
are nearly five thousand carriers of some sort in the US alone; clearly
“the network” is a terminological convenience that may be
destructively inaccurate as a reflection of the market environment.
Each of these carriers will evolve their component of the
composite public infrastructure according to the business and regulatory
pressures that carrier bears. Since
there is a fairly high level of interconnectivity and interdependence in
the service provider space, the individual decisions of major carriers
and the collective decisions of virtually all carriers will impact the
overall business and technology framework in which each must plan.
In short, it’s a chaotic, disorderly, process.
In the near term,
we can expect the service pressures created by the revenue base of the
present to influence incumbents in a connection-oriented direction.
It’s just good business. Regulations
will exaggerate this trend. Revolution
isn’t created by incumbents, it’s created by outsiders, and the
flight of capital from the industry will largely eliminate any of those
as market factors. Short-term,
then, the service model will drive infrastructure in a conservative
direction.
How about later on?
That’s where things will become interesting.
We can expect a wave of infrastructure build-out that will take
the public network from about one terabit per second of non-voice
aggregate traffic to about nine terabits in the next four years.
This build-out will be conducted predominately by incumbent
players obeying the service-to-network mantra of connection-orientedness.
In the five years
that follow, we’ll see non-voice traffic grow by at least 800%, and
virtually none of that traffic will represent extensions of existing
voice or connection-based data services.
The service framework of the network will have to shift, in the
period starting roughly mid-decade, to the connectionless framework of
IP.
There will clearly
be more to this shift than simply installing a few (or few thousand)
routers. The device
requirements of 2005 will be shaped not only by the service demand, but
by the network investment that’s already been made—and is still
substantially un-depreciated. Service
and network technologies like ATM, IP, MPLS, frame relay, transparent
LAN will all have to adapt to the shift.
The future is clearly set now as one of more gradual evolution
and many of the products and business paradigms of the present
(particularly those of startups) is linked to revolutionary paradigms.
These cannot succeed as long as they retain that revolutionary
flavor, and current players will have to adapt or sink.
Sinkings are
already increasing as the emerging (revolutionary) carriers fail, and
take with them the willingness to buy not for the transition period of
networking but for a hard-to-foresee set of future revenue paradigms.
The media and analyst community have conspired to feed the frenzy
of novelty in our industry, creating a glamorous forum so appealing that
it became as real (or more real) than the market.
That forum has now collapsed financially, and we’re left to
deal with the real world…the only one that was ever there at all.
Over the next two
years, the shape of the industry to come will be hammered out of the
disorder of the present, and we promise to help you all interpret as it
develops.
Well, where now? The prospects of near-term regulatory reform to jump-start the market now seem dim at best, so the industry now needs to consider what might happen to opportunities and spending in a market completely different from the one we’ve been conceptualizing in the past. That’s what we propose to do here, in a section that’s for subscribers only!
Netwatcher is a copyrighted information product of CIMI Corporation. Subscription is on a controlled circulation basis only. Subscription information may be obtained by clicking here.