TESTING THE SCALABILITY

OF DWDM NETWORKS

P. Ferreira1,2 and B. Cossa1

1:ITC Internet Telecoms & Convergence Consortium, MIT, USA

2: IN+ Center for Innovation, Technology and Policy Research, IST, PT

4th International Conference on Technology Policy and Innovation

Curitiba, Brazil, 28-31 August 2000
 MOTIVATION
Spurring demand for bandwidth ⇒ fiber exhaust

• Main challenges of today’s telecom networks:

– provide more and cheaper bandwidth
– integrate services using different technologies
Over the same physical infrastructure

• Forecasting demand:
– based on classical engineering formulas (e.g. Poisson)
– not factoring in the phenomenal growth of the industry:
• faxes, modems, teleconferencing, video on demand and e-business

• Fiber exhaust:
– significant lack of bandwidth capacity
– many IBPs have today 90% fiber usage (⇒ less protection, lower QoS)
 SOLUTIONS
DWDM claims to be the only true solution for the telecommunications industry

• To alleviate this problem:

– lay down more optical fiber cable. Not suitable for both:
• long distance carriers: networks exceed 10000 miles of fiber cable in the US
• small distance carriers: deal with gov. paperwork to deploy in metro areas

– increase bit rate using TDM. However,

• carriers have to purchase more capacity than what they initially need
• higher frequency ⇒ more fiber dispersion

– migrate to a Dense Wavelength Division Multiplexing (DWDM) system

• assigns incoming optical signals to specific frequencies
• multiplexes the resulting signals onto one fiber (⇒ scalability)
• success will depend on the intelligence of Optical Cross Connects (OxC)
• provides bit-rate and format independent wavelength routing (⇒ flexibility)
 A COST MODEL (I)
A generic IBP in the US with a full mesh SONET-based network

• Generic IBP in the US:

– 9 POPs, coast to coast
– 4 cells with full mesh network
– two links between nodes far apart
• Sizing the network:
– residential and business demand:
• bandwidth per user
• % of users active at peek
• internet penetration ratio
Network topology for the generic IBP considered
• market share
– incoming traffic from other IBPs through
NAPs (collocated with the POPs)
– demand assumed to grow 25% per year
 A COST MODEL (II)
DWDM POPs include OxCs and DWDM terminals, links have ER and OA

Legacy POP DWDM POP • POPs:

ISPs – legacy technology:
• ADM
ISPs
OC-48 • Router

OC-48
– DWDM technology:
local λ
• DWDM terminal
Router
• Optical Cross Connect (OxC)
local
Router
traffic • Router
OxC
long-haul λ
long-haul • Links:
ADM traffic
DWDM – Electrical regenerators
OC-192
(non-mux) OC-192 – Optical amplifiers
(up to 40 λ)
According to span rules
other POPs
other POPs

Generic Architecture for a POP in the legacy and DWDM networks

 RESULTS (I)
DWDM technologies cut significantly the amount of fiber needed

Capital costs over 5 years (LEGACY) Capital costs over 5 years (WDM)

27%
Fiber
7% Fiber 3%
5% In-line Amplifier
In-line Amplifier
13% 25% 37% Regenerator
9% Regenerator 74%
WDM
ADM
2% 10% Router
Router
59% 12% Cross-Connect

Breakdown of capital costs for the legacy and DWDM networks

• Multiplexing (40 channels in an OC-192) reduces significantly the

number of fibers deployed

– For instance, build a network to accommodate 3.6 Tbps costs 21 BUS$ using
legacy technology and only 6 BUS$ employing DWDM technology
 RESULTS (II)
Only DWDM encompasses significant economies of scale to meet terabit levels of demand

• DWDM network is more

expensive for low demand levels $14.00
LEGACY
$12.00 DWDM

CC per Kbps (in $US)

• Capital costs per Kbps served in $10.00
the long run: $8.00
– 5.5 US$ for the legacy network $6.00
– 0.5 US$ for the DWDM network
$4.00

$2.00
• Also from the NPV analysis:
$-
– the payback for both networks is Base x2 x4 x15 x30
(360 Gbps)
about 3 years Demand
– cumulative NPV decreases for Capital costs incurred per Kbps over time
the legacy network after 7 years
 RESULTS (III)
Only DWDM technology can meet high levels of demand growth

• The percentage gross margin

60%
reflects how good a company is LEGACY
at generating profit 50% DWDM

40%

• The percentage gross margin for

30%
an IBP with a legacy network
will decrease drastically if the 20%

demand growth rate is high 10%

0%
• However, for a DWDM 0% 5% 10% 15% 20% 25% 30%
backbone it remains constant Yearly growth rate of demand

even for high demand growth

rates Percentage Gross Margins for different
levels of yearly growth rate of demand
 RESULTS (IV)
Demand and fiber are the most important determinants for the capital costs

Business bandwidth 1000 100 1000 100

Business bandwidth
3200 9600
IRU 3200 9600 IRU
70% 50%
70% 50% Business Internet penetration
Business Internet penetration
250 50
250 50
Residential bandwidth
Residential bandwidth
20,000 95,000
Cross Connect
Residential Internet penetration 50% 25%

100,000 170,000
W D M (aggregated costs)
In-line Amplifier 66,667 83,333
250,000 320,000
OC-192 router port
Electrical Regenerator 79,500 90,000
50% 25%
Residential Internet penetration
OC-192 router port 250,000 320,000
200,000 250,000
Optical Amplifier
ADM 25,000 50,000 80,000 330,000
Electrical Regenerator
OC-48 router port 70,000 90,000 70,000 90,000
OC-48 router port
4 9 14 19 24 29 0 5 10 15 20 25 30 35
$ per Kbps (legacy network) $ per Kbps (DWDM network)

Sensitivity analysis of the capital cost per Kbps for both networks

• Demand parameters are the most significant to determine the capital costs
• The cost of fiber comes next (here represented by IRU - Indefeasible Right of Use)
• Third is the cost of equipment in the POPs, for the DWDM network, and the cost of the
equipment along the links, for the legacy network
 CONCLUSIONS (I)
DWDM technology will be a standard for every telecom infrastructure in the near future

• Technology Side:
– Capital costs in the POPs (⇒ DWDM cuts fiber significantly)
• 7% for the legacy network
• 74% for the DWDM network
– Capital costs per Kbps served in the long-run (⇒ DWDM is the only true
scalable solution)
• drop to 5.5 US$ for the legacy network
• drop to 0.5 US$ for the DWDM network
– Percentage gross margin (⇒ only DWDM can cope with high levels of
demand growth)
• for the DWDM network, it is fairly constant at 30% for demand growth rates
between 0% and 30%
• for the legacy network, it drops continuously as the demand growth rate
increases
 CONCLUSIONS (II)
The WAVELENGTH will be the unit of “thinking” in the telecom industry

• Policy Side:

– research incentives should promote:

• better forecast of demand levels
• intelligence of OxC towards all optical networks

– regulation and standardization:

• should be developed at the wavelength level

– bandwidth markets:
• higher flexibility ⇒ bandwidth on-demand and on-location