A Practical Navigator for the Internet Economy

Areas of Growth Will Be Custom Designed and Dynamic Enterprise Networks

Components of Public Internet So Intertwined that Shaky Equillibrium is Best that Can be Hoped For

Group of Experts Examines of Geography and Business Models and Begins to Understand that Network Design Issues Are Very Different and the Core, Edge and Last Mile

Click Here to get the Full Report in .PDF Format

QoS and the Economics of Networks

In this two month issue we continue our investigation into QoS and the economics of networks. We draw some large conclusions about the state of the globe. Most Carriers are trapped. They are faced with equally unpalatable choices. The first may be called "lose-if-you-win ". If they win as the result of being a survivor of consolidation, what they have won is the debt of the companies they take over. They will get more market share for sure , but the larger debt service will pare down their margins to the points where profits vanish. The other alternative, "lose-if-you-lose", sees them dissolved. Such are the economics of the global Best Effort Internet that with customer price sensitivity driving revenue in a competitive race to the bottom, it is likely that only a few will emerge as the result of major consolidation.

QoS is a viable technology but lacks the interoperability necessary to make it a significant force in the Public Internet. "Necessary cost cutting to survive" may have deprived many Service Providers of the skilled architects and engineers necessary to implement QoS. The emergence of QoS as a major service offering vector also is handicapped by the operators not seeing QoS as being able to provide any pricing differentiation for their services.

We conclude that anyone looking to understand the economic future of the industry must begin to look at global as well as regional assumptions about culture, technology policy and economics. Western players operate on different economic assumptions and model than their eastern counterparts. As the industry commoditizes and matures, the new markets are to be found in the East. This is an East unburdened by the "entropy of incumbency" (legacy assumptive thinking , business models, installed Cap-Ex sunken costs, regulatory morass and ambiguity, and debt of the West). Given these very different conditions the western technology mindset faces a significant handicap in dealing with India and China.

 

The talk given by CK Prahalad on May 6th at the University of Texas Dallas conference we attended illuminates the problem. Prahalad asked his audience to consider what it is that Western technologists design? Given a choice will they design a product? Or a more complex platform? Or a very complex entire system? He points out that normally in North America, they will opt for the complex system. He notes that over all, during the last three decades , the complex system designers have held sway. They have designed vast Greenfield systems like the Public Internet and the World Wide Web. For a while they built and people came, until, over burdened with cost and complexity and features that most users could never learn, the markets became satiated. As Moore's law combined with the passage of time to turn their complex products into commodities, the corporate infrastructure that was set up to support development of costly systems platforms found itself too top heavy with expenses to be able to support products that were both more simple and commoditized.

The Difference Between Western and Asian Markets

The IT and telecom companies of Europe and North America are locked in to a complex systems concept frame of mind from which they are unlikely to escape. It can be argued that the economies of scale of large companies using much capital to create complex and expensive content has created technologies that circumscribe consumer choice. The result has been market expectations either of high rates of return on the initial investment or for monopoly rents needed to payback the giant corporation and its myriads of investors. Of course a lot of the "over-build" wasn't done with a long term business model, but rather a "build it and sell it at a high short term rate of return on investment. Such a model thrives on large systems. Such large systems, be they the circuit-switched PSTN networks, the wide ranging fiber infrastructure of most ISPs, or Microsoft's OS, impose frameworks on users that limit their choices and force acceptance of a life circumscribed by the framework. They reward the forces that nourish the complex system and not the individual innovation that could challenge it. But within the time that ten generations of Moore's law become 20, the components of these complex systems have become commoditized, cheap in price and plentiful in number.

 

But unless they become monopolies, the complex systems don't evolve markets that can support their inevitable commoditization. Their markets become, instead, ever more specialized and segmented. These events lead to a spinning of 'corporate" wheels in efforts to design new systems with enough appeal to enough purchasers to provide enough income to enable a payback of amount of capital invested in the design. In reviewing both the marketing and profit and loss lessons of the last quarter of the last century, from automobiles to fine china to radio to agriculture, we find that the adoption of "one size fits all" economies of scale lead irreversibly to "production cost below sales price" commodities that either require governmental subsidies or an entire re-thinking of the system's goals. This rethinking of goals is necessary to meet the only alternative to the below cost commodity. The systems designer's only escape is to achieve market differentiation for their systems. Markets do differentiate along "perceived value" lines, only one factor of which is "lowest cost."

 

One meta-question is what kind of market is necessary to support what kind of system?

 

By comparison to Europe and North America, the voice and data markets in India, and China are blank slates. Here capital investment can start fresh, informed by the mistakes and possibly megalomania of the West. Here, fed by Open Source software, people are free of the Microsoft OS constraints. And they are also largely free of an Op-Ex costly and cumbersome wireline plant to support. Moore's Law and the resultant commoditization engendered by three decades of ever more mature integrated circuits creates in the East an economic environment conducive to design of new generations of electronics computers and telephones on a product level where cycles are short and capital invested not great. In such an environment, if a product doesn't succeed, there are plenty more resources to put behind alternatives which design and presentation are informed by the prior failures. The result is the emergence of new public networks will truly be portable and disposable.

 

Driven out of the U.S. by the collapse of the North American and European "Bubbles", engineers, trained in the West and experienced with the lessons learned, have returned home to India and China. There, on the South and East Asian mainland, they have joined with designers, and business strategists like CK Prahalad to fuel a very fresh approach to applying telecom to national economic development. These folk are getting very good at matching the economics that underlie the current cycle of technology capabilities with the ability to design and manufacture simple but durable products -not systems- that can be useful to and affordable by the local and generally very poor population. In part because they see a mass market numbering in the hundreds of millions, they are developing companies like Hua Wei and UT Starcom. These Firms, that by prior western standards would have been considered "fly-by-night", are positioning to make an onslaught against the older, more mature capital intensive western systems that (given the continued collapse of their mature markets) may not have the necessary underpinnings to survive. Consequently, with companies like Hua Wei, UT Starcom, and others believing that they can build commodity boxes to route the bits and devices to replace the phones and computers, it may be the case that we shall find the advanced telecom broadband networks that we thought we were building in North America only four years ago are now emerging from the ashes of NTT in Japan, full blown from government investment in Korea and nearly from scratch in Indian jungles and Chinese rice paddies.

 

Overall, the markets in China and India may be far more easily designed for and sold to than American and European markets. Western companies will start to build for them or die. What will look metaphorically like mesh networks can spring up like mushrooms in Asia, absent the need of central top down planning. The technology and economics of the technology markets are such that the Asia ground provides a fertile field for growth and exploitation. Intelligence is moving to the edges and the edges are found on the Asian mainland.

 

In the US what we will see is a stratification of networks that will exist according their users' ability to pay. The PSTN will be increasingly cheap, commodifiable and built on Asian models.

 

More expensive - but still cheaper than what the North American and European LECs could build them for- will be enterprise nets. The enterprises are presently discovering the huge amounts that they can save by no longer buying LEC based TDM services.

 

Those with the most immediate need for greatest reliability and security to protect their privacy and the integrity of their increasingly high- valued digital assets will be government, military and financial networks. These network classes will tend to cut across all national boundaries. Except for their own national security networks, government regulatory and legislative bodies, for the most part, will have to get out of the way, powerless to stop the deluge.

So What of the Business Models for Packet Networks?

Through their Best Effort Business model - send and send again until the packets arrive - packet networks have historically offered an inexpensive data transport that promised a lot for a minimal investment. While packet networks were in some ways more flexible than circuit switched forms of data transport, they were both less predictable and less easily controlled. More than merely cost effective, those features were designed in to the original Internet protocols to insure survivability and reliability in adverse circumstances or any localized perturbation. IP routing ports around both situations where a network link fails and in situations of intense congestion. In the 1990s, while the telcos stuck to more controllable PVC-enabled core networks with ATM and Frame Relay service guarantees for their more staid conservative clientele, by 1998, the nethead speculators were left to mob the fifty or so best effort backbones that had emerged in the speculative Bubble period. The shiny new best effort service may have had its draw backs, but it seemed to offer so much for so little for the applications in demand (email, web browsing, chat, newsgroups), that everyone felt compelled to commit to them some part of the enterprise's networking budget. Consequently, while Best Effort has become mainstream, the appropriate role for packet-based QoS and service guarantees is still debated. There is room for both but little agreement among the COOK Report panelists as to the likely overall size of any QoS market now or in the future.

Like Home Depot Do We Get Network Depot?

The telecom world's collision with IP has shattered the relatively small number of possibilities inherent in the old monolithic bell head networks. There are now dozens of technologies from hundreds of vendors offering thousand of products.

 

Anyone can go shopping for dark fiber, dim-fiber, other unused capacity, right-of-way and infrastructure to build almost any kind of network for any customer from the cornucopia of commoditized products developed in the 20 years since the break up of Ma Bell and the emergence of packetized "stupid" networks that swept the landscape in the final bubble years of the 1990s.

 

While the economic basis for huge dominant Public Internet is undoubtedly Best Effort, it is also losing money. Municipalities, and enterprises want service that they cannot get on the Best Effort public network. Entities that need new telecommunication resources must make investment decisions. We have been exploring whether the application of QoS tools to the public Best Effort networks can make a difference to the public internet or to other networks.

 

A bifurcation is beginning between first the "private-by-choice, mandate, or business need" enterprise, education, government and military networks, the increasingly expensive relative to alternatives PSTN, and second the "public-by-necessity" laissez-faire, increasingly insecure and privacy-invading Internet Commons for the general consumer, small business and SOHO sector. Under the conditions that are emerging, it looks increasingly as though the PSTN served by the LECs and by carriers is in perilous financial shape and will consequently become the ghetto of telecom.

 

The major aspect of the bifurcation is public versus private. The public packet networks are condemned to run under the Internet-enabled Best Effort business model where, absent a change in business model, those running the networks appear condemned to loose money for the foreseeable future until they either find a new business model or become consolidated and nationalized or acquired by the private networks.

Circuit Switched World

In the mid 1990s the telcos were offering ATM and Frame Relay on their data networks. These were technologies that would allow some service level commitments offering something beyond best effort: QoS being available in ATM, and to a lesser extent, Frame Relay networks. But from the beginning of standards efforts with RSVP and other early packet network QoS efforts, things did not go smoothly. The IETF process broke down across vendors with competing protocols and methods of resource reservation (e.g., adaptations of the End-to-End model of ATM, Edge-to-Edge Models, and hop-to-hop probabilistic models). The early attempts at QoS fell afoul of the lack of any viable Inter- Provider inter-working and Service Provider compensation mechanisms for taking the trouble to deliver the traffic of a competitor's customer to one of your own.

As our long previous discussions have demonstrated, QoS has relevance for the economics of the best effort networks. But right now, the relevance is limited by the fact that the Best Effort players simply do not see it as providing enough differentiation in their service to enable them to gain a positive ROI from investing money. This is attested to time and again in the pages of the symposium discussion in this issue. Where it remains far less expensive to provide another unused or under-utilized light path with carved out with simple-and-cheap-to- implement IPSec VPN virtual private lines, solving the delivery problem with available, already paid for inventory, the backbone networks are unlikely to embrace a more expensive QoS offering.

We conclude that for QoS to be able to make an economic impact in the public internet, QoS tools have to interoperate over a vast array of equipment. The interoperability is simply not yet there. Moreover, it doesn't look like it is coming there anytime soon.

But in looking at QoS anywhere, here also needs to be some assessment of what tools to apply to which kinds of situations, e.g., Stateful or Stateless.; and where in the protocol stack - Layer Two MPLS, Layer 1 MPLS, or layer 3 with flow routing?

What has emerged from our discussion is the insight that, for purposes of analysis, merely to view the Internet as a single indivisible entity is an error that masks solutions to the problem. We need to view "the network" or Internet as at least three inextricably intermingled and interactively interdependent components:

- the bandwidth rich backbone, the fiber rich middle, connecting (the "NFL" cities in the US and other fiber links connecting high traffic European cities with Tokyo, Hong Kong Singapore and perhaps Mumbai.)

- the Edge itself (defined as the distance from the first shared access point that the customer's traffic hits to the boundary at the CORE

- the customer access to the Edge switching and routing clouds (the local loop, from the customer premises to the first Edge access point)

At least four of our panelists (Frank Coluccio, Jere Retzer, Melissa Davis, Darin Wayrynen) have addressed these issues and shown that attempts to solve issues on any one or two, treating them as separate and self-contained system boundaries, have deleterious effects on the others. While MPLS is normally considered to be an Edge-to-Edge solution, Davis has recently explored its potential extension to the desktop. Coluccio, Davis, Retzer, and Wayrynen, in their own voices, have shifted our usually "Network-Centrist" perspective to one of the "network as utility like service provider to the applications that Enterprises need to be protected and accessed, and users either want or are required by business-need to access."

There is little doubt that if the QoS tools were mature, they could help improve the dismal economics that carriers face. But they are not mature. Caspian Routers using flow routing can increase the efficiency of a carrier's traffic, but how they can help that carrier deal with its edges and with exchanges with other internet backbones is not so clear. Axiowave using its technology again in a well-controlled well-bounded enterprise network can provide advantages.

Although the architecture and borders of QoS implementations remain so fuzzy in the public internet as to impede their near term adoption, we remain great fans of QoS in smaller private more specialized networks of which there will be more and more. But in the Best Effort Internet, the dominant view is still to over-provision and throw Cisco's new OC768 boxes in as needed. The only near term hope for the best effort Internet is serious consolidation. The fascinating Infranet approach to develop cross vendor network signaling capability (see pages 61-63 above) could be very important in the long run. The only problem here is that this is a standards kind of effort that may take several years.

The interesting news and innovation will come from the non-Best Effort private IP networks. These private networks will be the garden of QoS experimentation. Wireless is also an important wildcard in all of this - one to which we will turn in the coming months. While one should not expect to see any improvements from LECs or Cable-cos for the public internet, many significant municipal, government and enterprise networks are being built and are currently in use (Coluccio). Choosing the technologies for these that will offer the best return on investment requires a sharply- targeted, situation-specific analysis of the proposed business plan. Without a good knowledge of these global driving forces, the chance decreases for making the specific decisions that are best for whatever network one is planning. Individually interesting projects are being done. But it will likely be more than a decade before the global picture comes into better focus.

Contents

Economics of Long Haul Fiber Networks

Can QoS and Open Architecture Help Telecom Find Viable Business Model and Staunch Long Decline?

Until Telcos Unbundle as in Japan Users Are Unable to Take Advantage of Broadband Technology Since Network Owner Also Controls All Services

An Introduction to the September - October COOK Report p.1

Insight from India

We Built it and They Did Not Come

Designing and Marketing from the Customer's Point of View May Mark the Necessary Next Direction for Telecom p. 8

Sam Pitroda Father of Modern Indian Telecom Was First to Grasp Role of Telecommunications in Making India a Modern Society

His Work Founded on Intimate Grasp of Need to Appropriate Technology to Understanding of Society and Economy p. 13

Symposium Discussion

Debating the Economics of Best Effort and QoS Network Business Models

Dual Dichotomies of Public Versus Private Nets and Separation of Infrastructure from Services are Among a Myriad of Possibilities that Players Must Evaluate

 

Editor's Introduction p.18

The Korean Government as Anchor Tenant for the National Network p.18
The Problems of a Capital Repellant Business Model for Internet and Telecom. p. 29
Did it Work Because Government Ran a Command Economy? p.18
How Does Korean Experience Pertain to US? p. 19
Hewlett to Sell Wi-Fi Technology p. 20
The "Need" for Quality at the Edge - Are We Headed Back Toward ATM? p. 20
Korean Traffic Statistics [text box] p. 21
One Network "Type" or Two? p. 23

Why is Best Effort Better than it Used to Be? p. 23
No Inter-provider Roadmap to QoS Leaves a Big Problem for QoS p. 24
Down With So Called "Killer Aps" p. 25
Why the Network Is Not the Issue p. 27
Are Service Guarantees for Niche Markets or the Mainstream? p. 28

Mandates? Government or Otherwise? p. 30
Cost of Trans-Atlantic Bandwidth p. 30

Understanding National Networks - Assets or Not? p. 32
Many Networks - Ever Changing p. 33
Security Issues p. 34
Network Ownership and Attendant Business Models p. 35

Can a Business Model Encourage Usage? p. 37
Micro-payments? p. 38
Payment Clearing Houses p. 40
Do the Cell Phone Companies Have Better Economic Models? p. 42
Bandwidth is Not Treated Equal and Must be Carefully Defined p. 43
Global Crossing and QoS in Financial Industry p. 44

Separating Services from Infrastructure p. 44
A Wireline Public Domain? p. 45
ISPs Limiting of Users Spawns a Work-around Industry p. 46
Traffic Management in Oregon p. 47
Why ISPs Restrict Their Customers p. 49
How to Make the Network More Valuable p. 51

Economics of Giganews - Driving Prices lower p. 53
Where to Draw the Line Between Best Effort and QoS? p. 54

Economic and Technical Issues in Fiber to the Premises - Beware of Many kinks p. 55

The Deception of Fiber to the Premises p. 58

Network Signaling and Juniper's Infranet Initiative - We Return to Roadmaps and All That p. 60

Interviews and Discussion

PacketFront, a Swedish Company, Brings Working Business Model to Open Access Fiber Networks Technology Offers Web Based Network Operation and Customer Service and Billing Along With QoS Capable Edge Routers, p.64

Packet Front CEO Martin Thunman Explains How Company Evolved & Learned from its Predecessor B2, p. 73

Some Recent Developments in Regulation of No License Systems Smart Antennas and 3.65 gHz 24 W NOI Enhance Present and Future Capabilities, p. 75

Boucher's VoIP Bill Gives LECs Trojan Horse De-regulatory Bill Would End Common Carriage Requirements -- Legislation Would Hand the Local Phone Companies a Monopoly on all Physical Transmission Lines that Carry IP, p.78

Excerpted Highlights p.84

Executive Summary p.99

Contributors to this Issue

Affiliation given for purposes of identification - views expressed are those of the contributors alone

Frank Coluccio, President of DTI Consulting and designer of optical networks in New York City
Melissa Davis, optical network architect formerly with Cisco and now with RS Information Systems.
Beth Gage: Manager, Solutions Marketing, Juniper Networks
Vijay Gill, Director of peering AOL
Farooq Hussain, Network Architect Sprint, MCI, Currently Principal Network Conceptions
Ram Krishnan, Senior Director Technology, Axiowave Networks
Pete Kruckenberg, Architect Utah Education Network
Scott McCollough, Telecom Attorney Austin Texas
Andrew Odlyzko, Director Digital Technology Center, University of Minnesota
Dave O'Leary, Architect Juniper Networks
Taylor Reynolds, Policy Analyst, Strategy and Policy Unit, International Telecommunication Union
Jere Retzer Sr Mgr, Next Generation Networks, Oregon Health & Science University
Larry Roberts, Arpanet pioneer and CEO Caspian Networks
Darin Wayrynen, CTO GoodNet, Former VP Engineering Winstar, current ISP operator