Draft Draft Draft Draft Draft

A Gigabit National Data Grid/Fiber-to-the-Home (GNDG/FTTH) Initiative
for Computer Mediated Communications

www.rr.cs.cmu.edu/ndg/gndg-ftth.html

Raj Reddy (rr@cmu.edu)
June 18, 1997
This white paper was prepared for the use by
PITAC(President's Information Technology Advisory Committee)
as one view point of the Generation after the Next Generation Internet
.

Abstract

The main thesis of this paper is that the creation of a Gigabit National Data Grid with Fiber-to-the-Home is the only solution that permits a user to get gigabit data rates to home at affordable rates in the foreseeable future. This paper provides the rationale, cost models, revenue models, and action items in support of this proposal. Based on this analysis, it appears that the return on investment would be substantial even if only 10% of the Internet users were to subscribe to the GNDG/FTTH at costs much lower than the current T1 rates. Specifically, it appears that such a network can deliver gigabit-to-home service for something like the current cost of ISDN (about $50/month)!

Introduction

Creation of a separate National Data Grid with a Fiber-to-the-Home link is the key step that will enable computer-mediated communications. It is a desirable alternative to the current fragile and bandwidth-limited Internet. It sidesteps the slow response of the telecommunications industry to data communications needs. The Internet has grown 100%/year since 1988. Traffic has been growing 400%/year recently. This growth is creating a significant strain on the voice telecom infrastructure. This problem is likely to be exacerbated further by the high bandwidth access required by NetPCs and multi-media applications such as VideoPhone and VideoEmail. Other broadband applications of interest include entertainment-on-demand (unicasting), many-to-many conferencing (and collaboration), telemedicine, and distance learning.

Technology will soon allow us to build an independent National Data Grid of 100-gigabit bandwidth. This would be over 100 times faster than most current Internet backbone links. It has the potential to expand to 100 Terabits by the year 2010. It is the only way to provide gigabit data rates to individual subscribers at affordable rates in the foreseeable future.

A Gigabit National Data Grid (GNDG) appears to be an attractive business proposition, as well as an important initiative to ensure the nation's leadership. Such a grid could provide a GigaPOP in each of the 465 congressional districts in the country. Augmented by local fiber loops and spurs in metropolitan areas, over 90% of the homes and businesses in the country would be accessible within a mile of a GigaPOP (needs more careful analysis).

The GNDG/FTTH initiative is primarily intended for power users who are chafing at the current limitations and high costs of receiving high bandwidth data from the current telecom providers. However, it should be obvious to everyone that these power users will immediately start using the digital system for video telephony and entertainment delivered on an individualized basis. Only a small number of users will have this capability for many years to come, thus eliminating any immediate threat to established telecom and cable businesses.

The access connection from home to GNDG is expected to be conventional glass fiber. The current projections by Paul Shumate of Bellcore, suggest that that Fiber To The Home (FTTH) is competitive with other modes of providing high bandwidth to the home, taking into account the life cycle costs. Both a 100-gigabit backbone and FTTH are essential parts of a balanced solution capable of delivering the promised bandwidth end-to-end. Many of us have 100Mb Ethernet connections at work, but are unable to receive even 10% of the maximum bandwidth because of bottlenecks elsewhere in the Internet which is overlaid on the infrastructure created for voice telephony.

We propose that this network be limited for computer mediated communications. This essentially implies that the data into the home is digital and can only be accessed by users who have a computer or an information appliance which can accept digital data. Secondly, it assumes that the data is arriving at very high data rates and the appliance should have enough memory to be able to accept and buffer the data. The reason for insisting on this particular property is that it gets us immediately out of competing with the telephony and cable service providers who, for the foreseeable future, will be pumping analog data into the home.

The proposed architecture does not rule out other types of xDSL, Coax or wireless connections (both terrestrial and satellite based) to the GNDG backbone using existing infrastructures, albeit at lower data rates. At this time it appears that that the maximum data rates achievable by the existing infrastructures, without straining Physical Laws, are in the range of 10 to 100 Mb per second. It also appears that the cost of the electronics required for upgrading the current infrastructure at the CO plant and at home will be of the same order of magnitude as would be needed for interfacing fiber to home. The standards established by the FSAN (Full Service Access Network) Initiative would permit co-mingling of connections using different data rates.

The proposed GNDG/FTTH assumes that the data wire is independent of the telephone and cable wires going into homes. Most current proposals assume that a single wire will provide voice, video and data to home in the future. The telephone companies assume that this evolution will happen through xDSL and FTTx technologies, with FTTx being fiber to the curb initially. FTTC is appears to be at or below cost parity with copper in new or rebuild applications. The cable companies assume that this will happen through cable modems and digital TV technologies that provide significant broadcast bandwidth to be split up for unicasting of Internet data to homes. Terrestrial wireless companies such as Associated Communications and LEO satellite based solutions such as Teledesic also hope to provide digital access to home and mobile users. All of these assume evolutionary strategies based on current infrastructure and technologies. They also assume that each would dominate the others in providing such a service

The only argument could be which evolutionary path will be the least costly to the home. Most technologies being explored by telcos and cable companies appear to have maximum bandwidth limitation of about 100 Mbits per second. For a small percentage of the population, namely the Power Users of the Internet as well as Servers and Businesses, the fiber-to-the-home (FTTH) option with the opportunity to receive gigabits to home will be the preferred solution, provided one or more competing providers of GNDG/FTTH emerge in the marketplace. In the rest of this paper, we will describe the technologies, the costs, and the markets to further refine the proposed model.

The Technologies

The main technology driving the fiber bandwidth revolution is wavelength-division multiplexing (WDM). For more information see The National Transparent Optical Network Testbed : www.ntonc.org and Multi-wavelength Optical Networking Consortium: www.ito.arpa.mil/PIMeetings/BIT/MONET_Overview/MONET_Overview.html and www.bell-labs.com/project/MONET

Current commercially available solutions are capable of providing bandwidths of 40 gigabits per fiber (16 wavelengths at 2.5 gigabits each) and higher. Thus, leasing 5 fibers from one of the existing fiber owners such as Qwest or WorldCom, would provide a near term 100 gigabit data rate for a GNDG with an additional 100 gigabit as backup capacity. Assuming a new fiber network using a hundred strand fiber cable and technologies currently available in laboratories (see Bert Hui's writeup), one can expect to upgrade these data rates to around 100 terabits in a ten to fifteen year time frame.

The Costs

The cost of wiring the entire nation with a data grid of 50,000 miles appears to be in the range of $1 to $5 billion, depending on whose estimates we use. MCI, based on their cost structure for existing installations, estimates the cost to be in the $76,362 range per mile. Of this cost, the dominant cost is the fiber installation: build/construction cost and right of way management at $58,000. Other estimates put the cost at $15,000 to 30,000 (Jon Turner provides one such calculation at URL http://www.arl.wustl.edu/~jst/TransPrice.html., see also note from Peter Steenkiste), possibly because they ignore or assume much lower costs of construction and right of way costs! At $20,000/mile, the cost of a 50,000-mile national grid is a billion dollars. At $100,000/mile, which is the cost in dense metropolitan areas such as Manhattan, the cost is in the range of five billion dollars (See Appendix A for a more detailed model).

In addition to the National Grid, it is necessary to provide an access connection. According to Paul Shumate of Bellcore( see Prospects for FTTH ; see also his cost clarifications ), the cost of providing broadband access to the nearest GigaPOP with FTTH (using passive optical networks for facility and bandwidth sharing), would add a premium of $500 or less to the normal construction cost of switched access. This normal cost ranges from about $300 for some urban loops to over $3000 for rural loops, with the average being on the order of $1000. Thus, FTTH would cost $1500 on the average for those who require it as part of a new build and not share any of the existing wires. For the higher-cost rural loops, FTTH can be installed today at no premium. Based on the average figures and assuming FTTH technology, this capital cost depreciated over a 30-year period (typical today for facilities of this type) suggests a net cost of only a few hundred dollars per year for providing a high-bandwidth OC3 to OC48 connection to every home. This is 1,000 to 10,000 times more bandwidth to home, yet costs less than todayís ISDN rates!

Once you get the fiber to home, you also need to upgrade the computer system, for example by providing a Gigabit Ethernet chip on the mother board of the Intel P7 (expected to be available in 1998). Only the market demand will cause this to happen. Otherwise, we will need a separate add-on network card. Such cards are beginning to be commercially available today.

The Markets

Recent estimates put Internet use at about 20 million hosts used by about 35 million people. Zona Research puts the Internet market at $34.97 billion (see internet facts ). This includes Internet related equipment such as routers and modems, servers, software, hosting, services, and Internet access costs. This total market is expected to grow to $100 billion by the year 2000.

Of the 20 million hosts, approximately 2.5 million act as Internet servers, some of whom are large businesses with T1 or better connections. Assume that the access fee is fixed (say at 25/month -- the theme is "bandwidth is free" and the access fees are for recovering infrastructure cost) for all end-user connections to the National Grid. Even if only 10% of all the 20 million Internet hosts used the GNDG/FTTH and paid a $50/month flat fee ($25 for GNDG and $25 for the FTTH provider), the total monthly revenues would be $100 million each, for both GNDG and FTTH providers. That is a $1.2 billion return on a maximum of $5 billion investment for the GNDG infrastructure! (The Revenue Model given in Appendix B shows that, under various assumptions of monthly charges, market share and types of usage, the return on a five billion dollar investment could be 20 to 100% in the near term, growing substantially as the market penetration goes from 10% to 50%.)

Ultimate Impact of GNDG/FTTH Solution

Unlike other transitional technologies that are being proposed, fiber-to-the-home has the unique property of being future-proof. None of the others can overcome their inherent limitations while fiber is increasing its market share. Ultimately, the dream of digital convergence of a single wire carrying voice, video and data will come to pass. But the evolutionary path of how we get there could be very different than the current scenarios. (Needs further discussion here about the obsolescence of the old fiber already in the ground. Does this imply that all the existing fiber owners also have a huge capital investment to upgrade and to compete with any newcomers?)

Initially, the GNDG/FTTH solution should concentrate on providing Internet services for computer-mediated communications, leaving the current analog telephony and cable services alone. The reason for this should be obvious. Trying to solve the whole problem is likely to be time consuming and expensive in that additional interface cards in the PC for replacing the telephony and cable services are likely to be expensive for some time to come than leaving the existing solutions alone. Conversely, the current attempts at providing ADSL and cable modems for computer interfaces, are also turning out to be significantly more expensive because of the reverse installation and interfacing costs.

For many years to come, the GNDG/FTTH is likely to connect a few million Power Users. Many will be in businesses, on campuses, or working in government agencies. Some will be SOHO telecommuters. Among power users, one would use the video telephony enabled by the GNDG/FTTH. For most others, one would continue to use the POTS (plain, old, telephone service). As far as video goes, once the GNDG/FTTH is in place, we can expect some enterprising media service providers to begin unicasting entertainment-on-demand. While this will be slow in coming, it is inevitable. Most of the telecom and media giants will become the Western Unions of the 21st century, unless they are nimble and jump on the new bandwagon.

Action Items

How do we get there from here?

  1. Government's role: The Government should not engage in creating the GNDG/FTTH. That should be left to the market forces. However, the Government should act expeditiously to remove legal restrictions and regulatory inhibitors at federal, state and local levels. They should provide a small-royalty-based right-of-way access along the Interstates, and state and local highways. Various agencies of the government spend over a billion dollars a year on digital network services (it would be good to have a precise number), all of whom are likely to switch to GNDG/FTTH since the cost would be substantially lower!
  2. Telecom service provider's role: ATT, MCI, Sprint and several others already have the necessary rights-of-way and dark fiber in the ground to provide GNDG. However, they are probably unwilling or unable to forego variable prices and high cost ($600/month for a T1 line) based on old tariff structures unless competitive pressures push them in that direction. To be brutal, in a deregulated environment, only the nimble will survive.
  3. Information technology providers' role: Assuming that a gigabit to home will exist very soon, there are still no systems that exploit such data rates today. At best, they can handle 100 Mbps Ethernet. It is imperative that within a year or two, PCs come with Internet enabled interfaces, capable of accepting gigabit data rates. This requires higher bandwidth architectures and compatible operating systems and application software.
  4. Competition's role. Imagine that Microsoft, HP or IBM announces that they are setting up partnerships to create GNDG/FTTH. ATT, MCI, Sprint, Time Warner, TCI and the Baby Bells would soon have to announce competing HFC (Hybrid Fiber Coax) solutions. None of which would be able to offer OC3, OC12 and OC48 connections (that is, data rates over 155mb), unless they also replace their current analog wire by FTTH. In the latter case, they have no comparative advantage over any other provider of FTTH, leading to an interesting market situation where the ability to invest in huge capital infrastructures will determine the winners and losers.
  5. PITAC role: Other than disclosing such an option, PITAC cannot cause the GNDG/FTTH to happen. That is up to the commercial players.

Conclusion

Unless there is a significant flaw in the cost and revenue models, it appears likely that gigabit-to-the-home service will emerge in the near future. The most likely scenario will be that voice-telephony-service-providers and cable-service-providers will provide competing services at lower data rates and at lower costs. Permitting lower speed xDSL and coax connections to home in addition to the fiber option from the national data grid may result in a win/win situation. It will reduce the current stress on the voice telephony infrastructure, caused by computers connected by modems for hours on end, by rerouting all the data traffic to an independent data network, possibly using new area codes. However, it is difficult to see how the lower speed providers can compete and recover their investments in the long run, if the total cost of gigabit-to-the-home service cost is under $100/month!

 

Appendix A

GNDG/FTTH Infrastructure Cost Model
Raj Reddy
June 18, 1997

This document serves as an appendix to the GNDG/FTTH Proposal. It is an attempt to create a crude infrastructure cost model based on a number of assumptions, which are stated below.

GNDG/FTTH assumes the creation of a fiber grid following the main east/west and north/south interstates using 100-strand fiber cable. Since bits travel at light speed rather than automotive speeds, we donít have to have as dense a network of digital interstate highways. If we assume 5 east/west cables of 3,000 miles each and 8 north/south cables of 2,000 miles each, we are only talking about 31,000 miles of fiber. If we add metropolitan area network local loops of fiber of 20 cities with 50 miles and 80 cities with 20 miles, we have an additional 2,600 miles in high-density metropolitan areas. In addition, if we created spurs into every congressional district from the east/west and north/south grids so that there is at least one GIGAPOP in each district, the expected size of the grid is likely to be in the 40,000-mile range. The chart on the next page provides the cost of creating the gigabit national grid at different cost per mile and different sizes of the grid.

 

The following URL provide a map at different resolutions of what it would take to connect every congressional district using the right of way along the Interstate Highway System.

 

http://www.ul.cs.cmu.edu/rr/map.jpg

http://www.ul.cs.cmu.edu/rr/map1.jpg

http://www.ul.cs.cmu.edu/rr/map2.jpg

The following incomplete table attempts to create cost models per mile of installed fiber.

Cost per Mile

 

 

MCI *

ATT

Sprint

Qwest

WorldCom

           

LS Fiber

$7,161

?

?

?

?

50 mile regeneration

$3,452

       

Installation build/construction cost + ROW management

$58,000

       

Light gear/ switch end sites

$1,210

       

Light gear/ regeneration sites

$5,239

       

Switches - Nortel

$1,300

       
           

Total per mile

$76,362

       

* Thanks to Brian Mcmullan, Chris Stefano and Vint Cerf of MCI for providing these cost estimates.
For a full screen version of the following graph
click here

 

Appendix B

GNDG/FTTH Revenue Model
Raj Reddy
June 18, 1997

This document serves as an appendix to the GNDG/FTTH Proposal. It is an attempt to create a crude revenue model based on a number of assumptions, which are stated below.

No access fees (or revenues) are assumed from Baby Bells and Cable Companies in the following analysis, for providing lower speed broadband connection to/from the GNDG from/to home using their existing infrastructure.

Number of users of GNDG/FTTH

Total number of Internet users (in the United States) :

1999

30 million

2001

40 million

2003

50 million

 

Percent of the users in different categories:

Businesses already using T1

2.5%

Server hosts

10%

Hosts, non-server

87.5%

 

Expected market penetration by year and by category:

 

Businesses

Servers

Hosts

1999

50%

20%

10%

2001

75%

30%

15%

2003

90%

50%

30%

 

Flat access fees charged per month for access to GNDG/FTTH
Based either on data rates or Quality of Service or both

 

Businesses

Servers

Hosts

Low rates for OC3(155.5 Mb) users

$500

$200

$50

Medium rates for OC12(622 Mb) users

$1,000

$300

$100

Premium rates for OC48(2488 Mb) users

$2,000

$400

$200

 


For a full screen version of the following graph
click here