Transcript PPT

BUSINESS CONTRASTS BETWEEN HFC AND xDSL
Naval Ravikant
Manager, Strategic Planning
@Home Network
June, 1998
BROADBAND ACCESS LINES RACE TO THE FUTURE
Majority of Deployments will be Hybrid-Fiber Coax and Digital Subscriber Lines
• Cable deployments are leading the way
–
15,000-20,000 ADSL lines are installed in North America, whereas
over 200,000 cable modem lines are installed
Broadband Access Line Forecasts
9000
8000
Thousands of lines
7000
6000
5000
4000
3000
2000
1000
0
1996
1997
1998
1999
2000
2001
Year
ADSL
Source: America’s Network Magazine, IDC, Merrill Lynch
IDSL
HDSL
Cable
2002
TECHNOLOGY REVIEW: xDSL
DSL Bypasses LEC Switch and Uses Full 1MHz of Bandwidth
• Physical layer protocol for data-over copper
• Extends from subscriber premises to DSL Access Multiplexer (DSLAM)
located in Central Office. Bypasses LEC Switch, unlike ISDN
–
–
–
Filters in switch allow only 3-4kHz for voice, limiting analog modems
DSLAMs needed in every CO, whereas one analog modem bank can serve
entire LATA
Uses full 1MHz of bandwidth
• Unlike T1, doesn’t need repeaters every 3-6000 feet and causes less
bundle interference
• Discrete Multi-Tone line code breaks signal into 256 adaptive 4k chunks,
trading complexity and expense for reliability and rate-adaptivity vs.
Carrier-less Amplitude Modulation’s single 1MHz signal
• Asymmetric solutions send more aggregate data, taking advantage of
lower loop clustering and crosstalk at edge of network
Type
HDSL
SDSL
A / RADSL
VDSL
Range
12,000
12,000
18,000
1-4000
Up
Down
1.5Mb
1.5Mb
768kbps
768kbps
16-640kbps 1.5-6Mbps
1-2Mbps
13-52Mbps
Source: @Home study, America’s Network Magazine
Pairs
2
1
1
1
Application
T-1 replacement
SOHO T-1
Residential, passband
FTTC
TECHNOLOGY REVIEW: DATA-OVER-HFC
Modems Leverage “Fat” Coax Pipe to Set-Up Always-On Ethernet LAN
• Downstream programming begins at 50MHz, with 5MHz-42MHz reserved
for upstream communication. Each channel occupies 6MHz
• For data-over-cable, one channel in 50-750MHz range carries signals
downstream, and one in 5-42MHz carries signals upstream
• Downstream capacity is 36Mbps using 256 QAM and upstream is 10Mbps
using 16QAM or QPSK (more robust scheme is needed since signals
combine higher in the network, increasing vulnerability to interference)
• Bandwidth is shared, but since it does not need to be reserved (a la
dialup modems), overbooking ratios are very good - from 10-20:1
• Bursty nature of data traffic means that high-speed shared connections
are often preferable to mid-speed dedicated ones
• CableLabs tests show that performance degradation does not begin until
200-400 users per node are simultaneously accessing the network
Source: @Home study, America’s Network Magazine
BOTH TECHNOLOGIES ARE READY FOR PRIME TIME
Carrier Willingness to Invest in Deployment is the Real Constraint
• All major Cable MSOs have begun deployment of cable modem services,
most as part of @Home or Roadrunner, at prices ranging from $35/month
to $60/month
• 5 RBOCs and 19 CLECs have plans to begin deployment of xDSL this
year, at prices ranging from $60-$150 for residences and $100-1000 for
businesses
• Cable modem technology, provisioning, deployment and scaling work!
–
–
–
–
Over 200,000 full service paying customers
Security is a non-issue for residential subs. Encryption at layer 2-7 plus
insecurity of Internet limit issue
Shared bandwidth is not a problem with 1000 homes / node. (1000 homes x
10% penetration x 30% simultaneous online x 15% transferring data - tests
indicated 200-400 supported users / node)
Can add more channels or split node further as needed
• DSL technology has been tested in labs, trials, and HDSL deployments
–
–
–
–
HDSL already composes 60% of T-1 deployments
G.lite and UAWG addressing standards issues - DMT wins over CAP
Modems offered by ~30 vendors
Provisioning, loop testing and scaling issues still need to be addressed
BROADBAND TECHNOLOGIES FAVOR DIFFERENT MARKETS
DSL Naturally Favors Business Whereas HFC Naturally Favors Residential
DSL Attributes
Secure Loop
Dedicated loop bandwidth
Symmetric service (H, S)
Uses existing ISDN equipment (I)
Baseband (H, S, V)
Asymmetric Service (A, V)
Passband (A)
Favors
Business
Business
Business
Business
Business
Residential
Residential
HFC Attributes
Can integrate video / data signals
Costs decline with high penetration
Asymmetric Service (Telco return)
Symmetric Service (2-way)
Favors
Residential
Residential
Residential
Business
• Close to 60% of current T-1 lines are HDSL based
• Only 17% of carriers will deploy ADSL to residential users soon, and only 4%
will deploy HDSL to residences
• Fear of T-1 cannibalization and high frequency crosstalk will lead RBOCs to
choose a high price, low penetration service model
Source: IDC
COSTS: BROADBAND MODEMS
Silicon Follows Moore’s Law and Will be a Small Factor Over Time
–
–
DMT v. CAP debate settled - DMT
will lead deployments
Fast DSPs means DSL and 56k
modem can be implemented in
software on the same box
• Cable modems have 1-2 year head
start on price / performance curve
• Some DSL deployments (~20-30%
HDSL, 10% ADSL) will require
midspan units to bridge distance
limitations
Source: @Home Estimates
Broadband Modem Costs (pairs)
3000
2500
Cost per line ($)
• Cable MCNS standards initiative will
keep modems cheap and
interoperable
• Similar efforts underway on xDSL
side (MSDSL)
2000
1500
1000
500
0
1996
1997
1998
1999
2000
2001
Year
ADSL
HDSL
Cable Modems
COSTS: xDSL LOOP TESTING AND PROVISIONING
Reaching the Last 40% of Households Becomes Very Expensive
• About 40% of Homes are further than 18k feet, run through bridge taps or
loading coils, are served by DLCs, or are served out of a bundle carrying
a T1 signal
–
–
–
–
–
–
–
Installs are expensive - provisioning one circuit doesn’t tell you much about
adjacent circuits
Crosstalk between adjacent T-1 lines or other DSL lines causes interference
Line splices and loading coils that trap the signal above 4khz prevent DSL
Range limits of 12-18 kilofeet from Central Office. 25% of homes are outside
of this range in urban / suburban areas, up to 50% in rural areas
Current loop testing and provisioning systems only test the voice frequency
spectrum
25% of homes are served by DLCs, which have less space and power to hold
DSL equipment
$100-$150 per average install, $500+ for difficult installs
•
Example - US West charges $395 for average loop install
COSTS: DATA-OVER-CABLE PLANT UPGRADES
Rebuild Costs Vary by Market and Application
• Cable MSOs must upgrade plant to 2-way, with small node sizes
–
–
–
–
–
Entire neighborhood must be upgraded simultaneously
Opportunity cost of invested capital
Rural System upgrades are more expensive
Upgrade needed for digital TV & telephony
Marginal cost of adding a subscriber is low
Rebuild costs / home passed (avg. suburban system)
Homes passed / optical node
Fiber overlay on coax trunk
Optical / return-path amplifiers

300
$150
$5
600
$80
$4
Assumes no coax plant rebuild required to overcome channel lock
Source: @Home Estimates
1200
$40
$3
2000
$22
$2
BARRIERS TO MASS DATA-OVER-HFC DEPLOYMENTS
• Significant plant upgrades required
–
–
–
–
Short-term amplifier-only strategy can only support low take rates
Older systems need the most work
Telco-return implementations do not have a quick enough payback
Debt-load can make it difficult to raise cash - ROI expectations high
• Plant is also being upgraded for other services
–
–
–
Interactive TV revisited
Video channel capacity
Conventional, lifeline telephony
• Selling applications vs. pipes
–
Long term, people want IP dial-tone, but carriers are used to selling
applications and services, not connections
• Provisioning / Install needs to be easier
–
MCNS compatible, USB plug’n’play
• Lack of ubiquity
–
–
Constrains retail modem sales
Prevents mass deployment of packet telephony, tele-commuting, etc.
• Security and shared bandwidth “problems” are red herrings
–
–
–
Security available at higher layers
All bandwidth is shared at some point
More channels can be added (up to 4.5 Gbps capacity) and nodes split
BARRIERS TO MASS xDSL DEPLOYMENTS
• Complex provisioning and line certification issues
–
–
Bridge taps, loading coils, DLCs, range limitations, etc...
No testing systems in place
• Dinosaur distractions
–
–
–
Long distance telephony
Video aspirations
T-1, leased-line business
• Unrealistic expectations
–
–
Serving all customers
End-to-end ATM
• End result: Quarter speed ahead!
–
–
ISPs and CLECs will drive xDSL deployment (Covad, Northpoint, AIX, UUNET,
etc.)
Subloop must be unbundled - RBOCs will drag their heels
RBOCs WILL PREVENT MASS DSL DEPLOYMENT BY CLECs & ISPs
• RBOCs need to keep prices high to protect leased line and T-1 revenues
• Bury exorbitant markups in ATM circuit from CO to ISP facility
–
–
–
Force distance insensitive pricing even for short circuits
Forces cache to be in ISP facility rather than next to DSLAM
ISP has to pass cost through to customer
• CLECs and ISPs are relying on dry copper loops (used for alarm circuits,
etc.)
–
–
RBOCs are making these loops harder to obtain - raising prices, adding
loading coils, etc.
Dry loop now costs $30 per month instead of the $9-15 it would cost for an
alarm circuit. C.O.s to put the equipment in are strapped for space.
• 1996 Telecom Act lays foundation for unbundling, but is insufficient
–
–
–
Requires incumbent LECs to provide interconnection and mutually
compensatory rates for “exchange service.”
Unclear whether it is required for data services
Arbitration has focused on unbundling of entire loops, when it should focus
on subloop elements
•
•
•
•
Source: IDC
Network Interface Device
Distribution plant (2-4 wire copper loops running from customer to concentration equipment)
Concentration equipment (Aggregates dist plant circuits)
Feeder plant (Fiber / high capacity copper running from conc. Equipment to CO)
BASIC ECONOMIC COMPARISON
DSL’s Subscriber-Based Economics Win at Low Penetration Rates, but
HFC Triumphs When Plant Upgrade Cost is Spread over Large User Base
Annual cost per subscriber in Year 5 of Deployment
Annual Cost
$1,000
$800
$600
$400
$200
$0
1% 3% 5% 7% 9% 11% 13% 15%
Service Penetration in Year 5
Cable
•
•
•
•
ADSL
DSL modems @ $740, declining 30% annually; cable @ $250, decl. 20%
Cable upgrade is $55/HHP with $5 maintenance; 1/2 charged to data
DSL carriers don’t fix unreachable loops - 30% cost $400 to provision
Cost of capital = 15%; No differences in backhaul, CS, mkting costs
THE FUTURE OF CABLE AND DSL COMPETITION
• In areas where cable has already been rolled out:
–
Cable wins in residences (low marginal cost of adding subscribers)
• In areas where both services are deployed
–
–
Battle of marketing, services, and performance. The consumer wins
Cable has long term economic advantage for residences
• Areas where neither service will be rolled out or win:
–
Rural areas (wireless, satellite), FTTC (unlimited bandwidth)
• DSL’s chances improve if:
–
–
–
RBOCs do not try and serve all customers, and are willing to say “no” to those
served by poor loops
Subloop elements are un-bundled for ISPs and CLECs
Unrealistic requirements about network uptime and ATM are dropped
• Long term, the race will actually be for access capacity
–
–
–
–
Cable has 4.5 Gbps pipe, whereas xDSL is pushing limits of copper at 6Mbps
MSOs will push fiber out gradually as customers need capacity
RBOCs will replace aging copper with fiber wholesale, oversupplying some
areas and undersupplying others
RBOCs have more fiber in the ground, but MSOs are laying new fiber faster