A Practical Navigator for the Internet Economy

QoS is in the Eye of the Customer who Pays the Bills

Best Effort Built in Support of General Public Internet in an Anything Goes Environment

Group of Experts Continues Examination of Business Issues - Finds that the Smaller Network the Greater the Economic Accountability and the Larger the Potential Demand for QoS

Click Here to get the Full Report in .PDF Format

No Lasting Telecom Recovery Without Answers to these Issues

This three month issue explains why the prospects are dim for a full fledged recovery for telecom as long as the best effort paradigm remains as the only way of doing business.
The Internet became a capital repellant best-network paradox - see http://netparadox.com/ - because it was assumed by the implementers that a Best Effort network would be good enough as a foundation on which to build the new digital foundations of telecom. They were wrong. What resulted was a tragedy of the commons. There was every incentive for spam and for building peer-to-peer applications because resources were there that could be taken with being paid for. While, with fiber optic technology, links could be over provisioned, to sustain the best effort commons, everyone would ultimately have to over provision every link. With best effort paradigm of just-keep-on--trying-until-you-eventually-deliver-the-bits, everyone could plug into and use the highway without having to worry about paying with any kind of proportionality for their use.

 

Changing Network Paradigms

Let's look at networks in the 20 years since the break up of "Ma Bell." After the first decade, we had the circuit-switched, "gold-plated," five nine's reliable, SONET based, PSTN that had been optimized for voice over the duration of a century.

But suddenly, a decade ago, alongside that PSTN was placed the shiney new, efficient, decentralized-"telecosm" of the agnostic, packetized, best effort internet relying on a protocol that reroutes, drops off and tries again when the going gets tough.

During the resultant boom, almost all the new dollars got invested in giant commodity fiber-based players that were global in scope. Developments in optics and Ethernet created opportunities for data network fire hoses that were incompatible both with the straw to home and business local loops maintained by the telcos and the back-off-and-try-again politesse of TCP/IP.

The 90s saw upwards of a trillion dollars invested globally in a best effort commodity business that does "nice" things and serves hundreds of millions of people in ways heretofore unimaginable. The result is that the newly predominate global IP network infrastructure has been set up for best effort, Their offering has become massive amounts of packetized-bits. They have done a simply magnificent job of driving the price of bit transport below the cost of operations.

There are attempts to make use of routing protocols to provide classes of services. Generally these efforts run smack into the Procrustian beds of network boundaries and lack of settlements that incent other than best effort treatment across network boundaries. At the same time the best effort Internet has gotten very good at emulating telco protocols from SIP to softswitch and doing so in ways that suck the PSTN dry while not halting the downward plunge of the best effort network itself.

Consequently we are in the fifth year of a great sucking sound where it seems that all our resources are applied to the black hole of a perfect commodity.

Changes Appearing in the Midst of Network Chaos

The condition of the best effort network is certainly discouraging. We find that there is reason for some fresh hope because there are changes emergent. Fresh infrastructure is growing up largely unobserved alongside the best effort perfect commodity net. Namely the infrastructure of enterprises, government and military networks that, in the first case have to pay bills and generate profits for their sponsors, and, in the second case, must deliver mission critical services designed to take full advantage of monitoring huge streams of data in real time.

The "socialist commons" of the best effort Internet is being routed around. Governments, municipalities and global corporations are building their own telecom infrastructure as we speak. Tens of billions of dollars, and hundreds globally is going into outsource construction and operation contracts - moneys that used to go to carriers before those carriers got sucked into the singularity of the best effort, perfect network where the prime service differentiator has become cost cutting.
Carrier and service provider telecom is being transformed into role-your-own telecom, with hundred of thousands of lines making the switch with every passing month. Given the infrastructure costs and debt of the Carriers and LECs, it is very unlikely that they will be able to do much besides to continue to service their best effort constantly eroding base.

Eventually we will get to a new era of customer-provided telecom. When we do, it will be built on large scale reliable bandwidth needs and capabilities of the type that Melissa Davis and Jere Retzer have been discussing.

[Snip]

The technology is there the protocols are there and the dominoes are beginning to fall as the router and switch companies struggle to come up with products and protocols that can find the widest possible acceptance in these new narrow niche markets. What awaits is a shake out of the economics behind emerging markets for these new services - developments with which most of the industry is not aware.

We need to articulate the kinds of networks that can be set up to serve industry needs such as health care. We need to articulate how enterprises can coordinate their efforts to replace carriers or to transform those that seem capable of transformation. We need to spread awareness and understanding of government replacement networks that are being laid down.

We have three circus rings of network chaos and need a 200,000 foot overview and organizing premise to allow players to observe, articulate the pieces of and finally understand what is happening to the full system.

Is there an incentive to manage one's resources?

With these issues in mind we conclude that Best Effort versus Quality of Service discussions may make little sense unless carefully placed against a set of background assumptions. One of the most important assumptions is who pays? In everything except the Best Effort "commons," it seems there is an incentive to manage one's resources from a QoS point of view?

BEST EFFORT in the public Internet undoubtedly "works" even though, when one considers the number of players in the field under current conditions, it may be argued that any sound economic long term future is lacking. Short of consolidation, the cost versus opportunity of changing the way that things work at this Best Effort level presents huge problems.

But if you look at the impact of the public Internet's P2P and of the other "recreational" activities on corporate enterprise networks, you begin to find things like Packeteer popping up. As opposed to the global Internet for which no one is responsible, an enterprise network has boundaries and identifiable costs and incentives to keep network services running well within specified cost parameters. In other words the enterprise wants QoS.

But where then will the QoS come from? One solution is a traffic shaping effort in enterprises using things like Packeteer within enterprise boundaries to shape traffic largely at the application (layer 7) level. See http://www.packeteer.com The use here is to keep recreational web browsing and peer-to-peer traffic by employees from overwhelming bandwidth needed by corporate mission critical applications such as SAP. The principals are similar to those of carriers using their routers to constrict certain kinds of flows that Larry Roberts discussed in his interview earlier in this issue.

A different solution might be taken by carriers at layer two or three with the application of MPLS or Flow routing. The dilemma here would seem to be carrier understanding of mixed economic incentives. The carriers transit the enterprise packets over their WANs - along with all the best effort packets. One wonders how the carriers perceive any incentive to do anything about enterprise traffic flows in the midst of all the best effort tidal waves from their other customers? One hears about router and switch vendors come up with tools that facilitate carrier ability to do this? Perhaps there is yet another area where you have things like medical applications that may go from universities to labs, to hospitals and to Dr's offices. Do these become like "extranets"? There surely are opportunities for traffic management and QoS in such networks.

[Snip]

QoS is Being Seriously Explored

New companies like "Axiowave" have emerged from stealth phase with edge based, enterprise oriented routers that use flow routing and other QoS techniques. Most interestingly Larry Roberts has just gone off to form a new company called Anagran. It will do flow routing at the edge. Its market niche is QoS. Roberts is taking a very interesting gamble by developing an extension within IPv6 that will enable the establishment of signaling QoS where routers will talk to each other and be able to set the highest flow rates that the network can support at any point in time. We have acquired a copy of the extension specification and quote the scope and purpose on page 99 of this issue. We have found that the specification also deals with UDP packets in a way that will please potential VoIP customers.

It seems that Roberts has decided that he can author a spec in v6 in a way that he can implement it in hardware with more performance cost effectiveness than his inevitable competition. Relying upon v6 for the QoS appears to be a gamble for the North American Market. Europe is doing much more with v6 but even there carries are doing little. However an enterprise could run v6 with relatively little cost. Roberts clearly believes that give the increased security capabilities and the QoS dividends that the package will be sufficiently enticing for customers.

The general economic proposition is that a network, that is flow routed or is in other ways focused on QoS, can run at least two and perhaps more than 3 times as much traffic compared to way a best effort network can support without performance degradation. Average usage of best effort networks appears to fall in the 18 to 30 per cent range. Caspian claims that its flow routed networks can run at 80 to 85% capacity. The result is that on a given TCP/IP network with the same CapEx and possibly even less OpEx, from two to three times the amount of bits can be delivered at no increase in cost. Thus a provider could gain more income without increasing its expense. If that provider were an enterprise or a carrier it could exert better control over its network performance and save money.

This is a value proposition that we will continue to explore in coming months. Also emerging alongside this proposition are ideas of enterprise or specifically target extranets with dedicated QoS capabilities that can be put together independent of the public Internet. We expect to articulate some possible experiments in this area - especially in the health care area. In the meantime we will close with Larry Roberts' articulation of the issue on June 13.

"We clearly have created the myth that Gordon suggests that many "net heads" think TCP/IP must be "best effort" only. However there is absolutely no basis for this myth today. One network is clearly cheaper to operate and to mix traffic types is clearly beneficial in that we can fill the holes with best effort traffic. The cost of a bit with low delay, low loss, and guaranteed bandwidth is not much higher than a bit of best effort traffic, it is only scheduled better and thus a pipe cannot be filled if that is all there is in traffic. That is, video conferencing alone on a network must leave some unused space. If one does best effort alone on a network, it also does not fill the pipe since the sources are very bursty and fractal traffic typically would have bad loss and delay if the pipe was always filled. But mixing the two type, or all types, is actually better for the network, it provides some stable traffic streams where bandwidth is demanded at the time requested, and it provides some file traffic which can be moved around in time to fill the holes.

[Snip]

Second, I see no reason to presume that any current IP carrier could not upgrade to offer this combined service. They may have to buy some (not all) new network equipment to support IP with real QoS, but the current fiber, pops, DWDM, organization, and best effort routers can clearly be used in creating the new mixed network. There is an analogy here to the Internet software boom where people thought they could put current well run businesses like grocery stores out of business by starting over on the Internet. The old stores won.

In some cases, new businesses occurred like Ebay and Google where there was no prior business. But where the business existed, they could restructure and do the new thing. IP networking is clearly not going to change much just because someone adds a new router that supports QoS. The big, well run companies already doing IP can easily move into the market if it is proven. So I doubt we will see a transition to new small QoS only or application only carriers. As IP QoS is proven, it will be mixed with the current traffic on the current networks, adding a new higher revenue per bit service, but not hugely higher since the cost is so close. That still leaves the carriers as bit movers, just like grocery stores are food movers. But the [bit moving] business is needed and will eventually be profitable as the overbuild gets used and the situation settles down.

I do not believe ATM or TDM were ever economic competitors to IP, no matter what QoS is desired. IP is far more economic due to inband signaling, variable packet size and variable bandwidth. But IP has been burdened with the myth that Gordon mentions that it only does best effort. The only reason for this myth was that memory to do QoS was too expensive in the 70's and 80's. That clearly is not the case today."

Larry has issued a provocative challenge to our symposium. In mid July we will offer an up date plus several fresh looks at how to think about and understand the identity of one's customer.

Contents

Economic Pressure on Long Haul Fiber

Five Years After Bubble Burst Prices Have Plunged Yet Nothing Fundamental Has Been Fixed

Examination of Data Network Woes Shows Termite-Riddled Foundation Leading to More Bankruptcies in Absence of Broader Understanding p. 1

Backbone IP Networks: Why the Hierarchical Peering Model Is Broken

Like Electric Grid - No One Wants to Pay for Connecting
as Fiber Glut Overturns Dominant Position of Tier Ones p. 8

Some of the Business Model Issues of Peering and Interconnection Or Why the Government Could Eventually Be Forced to Take Over p. 21

We Are Now a Decade into IP Based Optical telecom without a Viable Business Model -- An IP and Optical Network Architect Looks at the Collective Insanity of the Bubble and Calls for Fresh Assessment p. 25

Symposium Discussion

We Built it But They Did Not Come - Now What?

Broken Model of Peering and Interconnection and Fiber Over Supply Not Being Solved In Absence of Viable Business Model - Global Foundation of Digital Telecom Still Termite Ridden p. 28

Houston: We Have a Problem Editor's Introduction, p. 28
The Problems of a Capital Repellant Business Model for Internet and Telecom. p. 29

Chapter One -Transition Problems Inevitable: p. 30

Larry Roberts: The Packets Do Not Say Who's Responsible p. 33
US May Take Another Five Years to Dig Out of FCC & Boom Created Problems p. 34

Chapter Two - Best Effort Paradox & Need for Initiatives Other than Best Effort Editor's Introduction: p. 36

Problem Solving Options: (a) UN, (b) MPLS, (c) Internet IXs? p. 38
MPLS and Global Crossing - Problems and Prospects, p.39
Dynamics of Disappearance of Settlement Free Peering, p. 42
Price Collapse Makes Avoidance of OpEx More Important p. 44

Chapter 3: Is There a Point of Price Stability? p. 46

R and E Owned Fiber User Controlled Light Paths and IPv6, p. 47
Where Things are with VoIP, p. 48
Searching for a VoIP Revenue Model p.49
Summary: Don't Look for Quick Fixes from IPv6, UCLP or MPLS p. 52
MPLS VPNs and MPLS IXs p. 53
Lambda Switching and OpEx Costs, p. 54
Exchange Point Alliance? p. 56
A Challenge From Melissa Davis - Enable Applications that Generate User Demand and Higher Revenues p. 58
Market for Faster-than-Real-time File Transfers? p. 59

Chapter 4: Some Possible Economic Impacts of Not having QoS p. 60

Chicken and Egg - No Incentive for Customers to Conclude that Best Effort is Not Good Enough, p. 61
Meanwhile Verizon Buys Fiber, p. 64
Issues Involved in Quality of Service Business Cases, p. 65
Best Effort Versus QoS Revisited, p. 67
MCI Emerging from Bankruptcy Intact, p. 70
Interprovider MPLS p. 71
So Where are We? A Summing Up, p. 72
Is Best Effort Leading Us to a Bad End? p. 76

Chapter 5: A QoS Business Model: p. 78

But for the Masses Best Effort Seems Good Enough - Cogent? p. 81
Don't Bet on Cogent p. 83
Telecom Bandwidth Provisioning Trapped in a Deflationary Spiral? p. 85
Unless We Have Profits p. 88

Exploring the State of Long Haul Packet Networks

A Survey of Global Business Conditions in a QoS Context of IPv6, MPLS and Flow Routing -- A Conversation with Caspian's Larry Roberts, p. 90

Anagran Inc. - New Larry Roberts' IPv6 Qos Signaling Extension Key to Understanding His COOK Report Interview p. 98

"Bookend" to Network Economics - A Summary and an Anticipation -
by Melissa Davis p. 99

Chapter 6
In Search of a Sustainable Business Model
Are Stupid Networks Less Secure and Less Profitable? p.100

Marketplace Viability of IPv6 Rejected p.100
Flow Routing p.101
Larry Roberts Responds to Comments p.102
Some Actual Limited Use of QoS Tools p.104
Performance Standards and Backbone Traffic p.106
Some Figures on the Total Size of the Industry and Backbone Economics p.107
Backbone Economics and Comments on Korean Traffic p.108
Europe May Not Be in Wonderful Shape Either p.109
QoS, Internet2 and Damaged Goods p.110
Neither Best Effort, Nor Telco - Are Tools Cheap Enough to Begin to Build Industry Specific QoS Overlays? p.112

Chapter 7
Best Effort or QoS or
Some Combination of the Two?

Stuck in the Best Effort Mind Set p.115
A New Found Relativity - Some Routing Protocol Disagreements and Business Models p.118
Signs of Emergence of New Non Best Effort Enterprise and Government Private Networks p.121
Economic and Security Administrative Issues of Build Your Own p.121
Economic Administrative Complexity Underlying Building Your Own Fiber Infrastructure p.122
Roll Your Own Or Lease Some Wave Lengths? p.123
More Examples of Building Your Own Networks & Bell South VoIP Trials Not on Internet Backbone p.124
The Network Is Not the Issue p.125
Korea and a Gigabit to the Doorstep p.127
An ITU Korean Study Summarized p.128
Selling IP- VPN Services p.130
Roberts: One Network Can Do Both BE and QoS p.132

New America's Cartoon Guide to Spectrum Policy Reveals Shortcomings of Economic and Policy Language in Coping with Radio Design p.134


Guide Does Not Deal Adequately With Subtleties of Part 15 Rules p.134
Difficulties of Deciding How Much Power to Allow p.134
Underlays or Overlays in the Broadcast Bands? p.136
Property Rights p.138
Property and Property Right Not the Same p.140
Technology Has Negated Monopoly Spectrum Rights p.143
It's Architecture of a Radio System not the Individual Power p.148

Dave Hughes Explains Sip, VoIP and FW Dialup in Voice Communication with Nepal over the Internet p. 154

Digital Photography Comes of Age -- Two New O'Reilly Books Reviewed in the Context of New Tools p. 156

Highlights p.160

Executive Summary p.181

 

Side bars:

Over Provisioning Dilemma: Why Adding Bandwidth at One Link Just Shoves Congestion Elsewhere p. 38
Technology and Investment Have Out Run Demand, p. 63
Transatlantic Virtual Peering, p. 69

An Example of the New QoS
Network Business Model p.115 -116

Contributors to this Issue

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

Falk Bornstaedt, Vice President, Product Management IP/Data, T-Systems,
John Curran, former CTO XO, principal at Istaff
Peter Cohen, Peering Director Telia-sonera
Frank Coluccio, President DTI Consulting -designer optical networks in New York City
Melissa Davis, optical network architect formerly with Cisco- now with RS Information Systems - and valuable Editor for this issue
Alex Goldman, Reporter editor www.isp-planet.com for Jupitermedia, covering ISP issues
Dan Golding, Senior Analyst Network and Telecommunications Strategies Burton Group
Roxane Googin, financial analyst Editor Hi Tech Observer
Steve Heap, CTO of Arbinet a bandwidth broker and VoIP Traffic terminator,
Farooq Hussain, Network Architect Sprint, MCI, now Principal Network Conceptions
Peter Juffernholz, former Peering Director Teleglobe, currently developer of Business Alliances DT
Pete Kruckenberg, Architect Utah Education Network
Martin Levy, former VP engineering Concentric, independent consultant
Tony Li, Router Architect at Cisco, Juniper and Procket, currently Consultant
Francois Menard, Canadian policy expert and municipal fiber network architect
Don Mitchell, former Program Manager National Science Foundation currently retired
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
Dave Siegel, VP Backbone Engineering, Global Crossing
Max Smetannikov, Director Business Development Global Analysis Services in NYC
Jim Southworth, former chair DSL forum, VP Concentric, now Secure Pathways CEO
Bill St Arnaud, Director CA*Net4
Darin Wayrynen, Fouder Goodnet, former CTO Winstar, current WISP operator