Why Cable Bends Matter in
Enterprise Networks and Why
Multimode Fiber Prevails
Sharon Bois
Corning Optical Fiber
May 22 2010
Agenda
• Multimode fiber remains the most cost-effective
choice for enterprise networks
– Multimode primer (classification and bandwidth)
– Benefits of multimode fiber (versus single-mode fiber
and copper)
• Next generation multimode fibers and standards
– OM4 and next generation speeds (16 Gb/s, 40 Gb/s
and 100 Gb/s)
– Bend-insensitive multimode fiber
Multimode fiber types classified
based on bandwidth values
“OM”
Designation
Core Diameter
(µm)
EMB
(MHz.km)
OFL 850/1300
(MHz.km)
10G Link
Length
OM1
62.5
-
200/500
33 m
OM2
50
-
500/500
82 m
OM3
50
2000
1500/500
300 m
OM4
50
4700
3500/500
550 m
• Optical Multimode (OM) designations are per ISO/IEC
11801
• EMB = Effective Modal Bandwidth (Laser BW)
• OFL = Overfilled Bandwidth (Legacy/LED BW)
Lasers require new bandwidth
measurement systems
OFL (Overfilled Launch)
• Designed to predict performance of lowspeed LEDs, not lasers
• Power distributed over 100% of the fiber
core, like LEDs
• Perturbations in index profile undetected
Light Sources
(Typically 10 and 100 Mb/s)
EMB (Effective Modal Bandwidth)
• DMD (Differential Mode Delay) based
measurement
• minEMBc or DMD-mask
• Power distributed in a narrow region
• Simulates an actual laser launch
• More accurate indication of performance in
high-speed laser-based systems
(1, 2, 4, 8, 10 Gb/s and higher)
Characterization Methods
DMD (differential mode delay) Mask
Fiber Core
1 of 2 DMD-based standards compliant measurements
One laser type scanned across core
BW defined by most delayed pulse
DMD output is “Normalized”
6 Masks Applied for OM3 (3 masks for OM4)
Must only pass 1 mask to be OM3 (or OM4) compliant
Fiber Core
≈5µm
Laser
TSlow
TFast
25%
1-2 µm
Laser
Laser
Laser
Laser
Pass = OM3 (2000 MHz.km EMB) or OM4 (4700 MHz.km EMB)
Fail = OM2 (< 2000 MHz.km EMB)
Tdelay
Characterization Methods
minEMBc (min Effective Modal BW – calc)
Fiber Core
1 of 2 DMD-based standards compliant measurements
Simulates several laser types scanned across core
BW defined by most delayed pulse
Fiber Core
≈5µm
e.g.
e.g.VCSEL
VCSEL#5
#5
TFast Bandwidth
Bandwidthvalue
value
==3128
3128MHz.km
MHz.km
TSlow
Laser
e.g.
e.g.VCSEL
VCSEL#3
#3
Bandwidth
Bandwidthvalue
value
==2563
2563MHz.km
MHz.km
Laser
Laser
e.g.
e.g.VCSEL
VCSEL#1
#1
Bandwidth
Bandwidthvalue
value
2137
MHz.km
===2137
2137MHz.km
MHz.km
Laser
Laser
“Hot inside” laser
“Mid-range” laser
“Hot outside” laser
T delay
minEMBc Value
Note: BW values provided for illustrations purposes only, drawing not scale
Different laser
characteristics simulated
Multimode fiber dominates in risers &
data centers
Horizontal
• Predominately Copper
• 10/100/1000 Mb/s
• Zone (FTTE) fiber
growing
In-Building Backbone
• 80% fiber and increasing
• 35% 1 Gb/s - 65% 100
Mb/s
• Multimode fiber
dominates, OM3 preferred
Campus Backbone
Data Centers
• 60% fiber and increasing
• 1, 2, 4, 8 and 10 Gb/s
• Multimode fiber dominates,
OM3 strongly preferred
Source: Corning Analysis
• 95% fiber and increasing
• 10 Gb/s initial deployments
• 70% 1Gb/s - 25% 100Mb/s
• Fiber preferred, singlemode fiber continues to
gain
Historically, there have been many
benefits of fiber vs. copper
Performance
•
Pathways and
space
– Smaller, lighter
cables
ƒ Less cable fuel
load
– Easier installation
Fiber
– Higher data rates/
longer link lengths
– Low latency
– Network security
– Immune to EMI,
RFI and cross-talk
– Longer cable life
OM4
OM3
OM2
10 Gb/s Link Lengths at 850 nm
OM1
0
100
200
300
400
Distance (m)
500
600
Cat7
Copper
•
Cat6a
Cat6
10GBASE-T Link Lengths
Cat5
0
100
200
300
400
Distance (m)
500
600
With 10G the list of benefits of fiber
vs. copper is growing
• Electronics port
density, power and
cooling efficiencies
= GREEN
10 Gb/s Operating Cost
Fiber vs. Copper
Fiber
~1-4 W
Copper
Power Consumption
~8-10W
Cooling Requirements
• Cost position
changing with 10G
– TIA Fiber Optic LAN
Cost Model
– Complexities of 10G
copper testing
Transceiver Size
Data Center Area
$
$$$$
Active components costs dominate
enterprise link economics
• Fixed costs ~ the
same regardless of
fiber choice
Fiber Optic Cable
Transceivers
24%
Jumpers,
Connectors
< 1%
• Transceiver costs ~
¼ of total system
costs
– Key area for savings
with multimode fiber
– Greatly outweighs
difference in singlemode versus
multimode cable cost
1%
Patch Panel, Rack
< 1%
Fixed Cost
Switch Electronics
74%
Typical 300 meter backbone
Source: www.foundry.com, www.peppm.org, Corning analysis
Multimode fiber solution saves ~50%
over single-mode fiber solution
Assumptions
OM3 fiber
– Supports 10 Gb/s over
300m
– Lowest cost upgrade path
to 10G
0.5
OM3
OM4
1300 nm
1.5
1
Tx/Rx
Small ∆ for
10G capability
1300 nm
2
Hardware
Single-mode
Fiber Cable
Key findings:
– Cable very small portion of
link costs
– MMF 850 nm (SX)
solutions always lower cost
•
2.5
850 nm
•
Relative System Costs: 1 Gigabit over 300m
850 nm
– 300 m, 24 fiber count cable
– 24 fiber Passive
Interconnect (x2)
– 18 x 1 Gb/s Transceivers
850 nm
•
0
OM2
OM-2 (LX) OS-2 (LX)
LX = Long Wavelength = 1300 nm
850 nm continues to provide cost benefit at 10G and beyond
850 nm continues to provide cost
benefit at 10G and beyond
•
850 nm VCSELs ~90% of
optical 1G enterprise market
850 nm 10G VCSELs just
entering high-volume
manufacturing
– 850 nm continues to be low-cost
solution for 10 Gb/s
– Low cost solutions for 100 Gb/s
have been identified
– SFP+ 850nm transceivers
continue to drive price down
•
LR (1300 nm) solutions may
capture some market share in
enterprise networks
– Small percent of new installs
10 Gb/s Transceivers
3.5
850nm
3.0
1300nm
Relative Cost
•
2.5
2.0
1.5
1.0
0.5
0.0
2003
2005
2007
2009
2011
Time
Source: Corning estimates
North American market now majority
50 µm
Multimode Fiber Market Demand
70%
65%
60%
55%
50%
45%
40%
35%
30%
Q1 2005
50 µm
Q1 2006
Q1 2007
62.5 µm
Q1 2008
Q1 2009
Q1 2010
Source: Burroughs Report
More 50 µm sales than 62.5 µm sales since Q1 2008
Laser-Optimized 50 µm continues to
grow
50 micron Market Demand
65%
60%
55%
50%
45%
40%
35%
OM2
OM3/OM4
30%
Q1 2005
Q1 2006
Q1 2007
Q1 2008
Q1 2009
Q1 2010
Source: Burroughs Report
OM3 has been majority of 50 µm since Q1 2007
Multimode fiber preferred for
enterprise networks
Single-mode
Information carrying capacity
X
Distance supported
X
Fiber price
X
Multimode
Connector price
X
Transceiver price
X
Ease of handling (core size)
X
Multimode = Lowest price system for shorter links (10G to 600m) with many
terminations
Single-mode = Best solution for high bandwidth longer links (10G > 600m)
TIA-942 recommends the use of laser-optimized 50 µm multimode fiber for
backbone cabling because of its capability of supporting higher network speeds
over longer distances while being more cost-effective to implement than singlemode fiber.
Part II
• Next generation multimode fibers and
standards
– OM4 and next generation speeds (16 Gb/s, 40 Gb/s
and 100 Gb/s)
– Bend-insensitive multimode fiber
OM4 standard approved by TIA
in August, 2009
•
OM4 is 50 µm fiber with higher effective modal bandwidth than OM3
– Extra bandwidth can be used for higher bit rates, longer link lengths or
increased margin for more connectivity
•
Existing “OM” designations (per ISO/IEC 11801) are shown in the table
below
•
IEC proposal for OM4 has yet to be approved but highly likely it will be
harmonized with TIA
“OM”
Type
Core
Diameter
(µm)
EMB
(MHz.km)
OFL 850/1300
(MHz.km)
10 G Link
Length
100 G
Link
Length
OM1
62.5
-
200/500
33 m
-
OM2
50
-
500/500
82 m
-
OM3
50
2000
1500/500
300 m
100 m
OM4
50
4700
3500/500
550 m
150 m
Parallel optics are preferred for
multimode fiber objectives
40 Gb/s
•
4 fibers x 10 Gb/s for transmit
•
4 fibers x 10 Gb/s for receive
•
One 12 fiber ribbon
100 Gb/s
•
10 fibers x 10 Gb/s for transmit
•
10 fibers x 10 Gb/s for receive
•
Two 12 fiber ribbons
IEEE approves 40G/100G standard
•
The standard supports 40 Gb/s over:
–
–
–
–
–
•
At least 10km on single-mode fiber
At least 100m on OM3 MMF
At least 150m on OM4 MMF
At least 7m over a copper cable assembly
At least 1m over a backplane
The standard supports 100 Gb/s over:
–
–
–
–
–
At least 40km on single-mode fiber
At least 10km on single-mode fiber
At least 100m on OM3 MMF
At least 150m on OM4 MMF
At least 7m over a copper cable assembly
OM3 100 meter distance allows for 1.5 dB of connector loss
OM4 150 meter distance allows for 1.0 dB of connector loss
Value proposition for OM4 depends
on application
Applications
System Operating Link Length vs Laser Bandwidth
10G Ethernet
500
400
OM4 Benefit
Link length (m)
600
300
200
100
0
2000
•
•
3000
3500
4000
Laser Bandwidth EMB (MHz.km)
4500
4G Fibre Channel
5000
OM4
OM3
Significant value for OM4 at 10G Ethernet
Little value for OM4 at 4G regardless of EMB value
–
•
•
2500
40/100G Ethernet
16 G Fibre Channel
Dispersion limited because of broad spectral width
16G has tighter spectral width than 4G so value increases
Although 40G/100G is based on 10G arrays, looser specifications for 40G/100G
transceiver arrays significantly reduce the value
OM4 at 40G/100G extends cost
effective MMF solution
80%
60%
40%
20%
•
Cumulative Frequency
Relative Frequency
100%
– Reducing connector loss to
same level as OM4 allows
OM3 to support 120 m
•
Extending OM4 distance to
150 m with existing
transceivers covers ~ 90%
of data center links
•
OM3 and OM4 fibers can
support even longer
distances, but transceiver
spec change is required
0%
0
50
100 150 200
Cable Length (m)
250
Length Distribution Cumulative Frequency
Source: Corning Cable Systems
Objective of at least 100 m
on OM3 covers ~ 70% of
data center links
Moves, adds and changes (MACs) can
cause a structured cabling system to
look more like a rats nest
However…
Initial installations that follow
bend radius guides and
structured cabling paths don’t
have to worry about signal loss
due to inappropriate bends
Over time, MACs lead to mismanaged cabling resulting in:
• Congestion in sub-floor space
• Bend-induced attenuation
• Restricted air flow
• Negative impact on cooling efficiency
Data center “wish list” points to
need for effective cable management
•
•
•
Increase density of factoryterminated solutions
Improve slack management
•
Relieve congestion in pathways
and spaces
Improve airflow
•
•
Eliminate polarity concerns
Improve MACs
Poorly installed cabling can degrade
network performance
•
•
•
Element
Percentage of
Overall Cost
Expected
Lifespan
SOFTWARE
40%
5 years
COMPUTERS
30%
3 years
SERVERS
20%
3 years
NETWORK
CABLING
10%
15 years
Source: Datalan-Network-Infrastructures
•
Cabling is a relatively small
fraction of the initial network
spend
Cabling has the longest
expected lifetime of the major
network elements
The potential for network failure
due to poor cabling is high
Cabling is often an “afterthought”
but it shouldn’t be
– Key to ensure that the cabling
won’t become the most
expensive part of the network
Fundamentals of macrobending
in multimode fiber
•
Multimode fiber has many modes of light traveling through the core
•
As each of these modes moves closer to the edge of the core it is
more likely to escape, especially if the fiber is bent
•
In a traditional multimode fiber, as the bend radius is decreased, the
amount of light that leaks out of the core increases
Core
Cladding
Dissipation of
energy
Bend-insensitive multimode fiber
prevents light from escaping
•
A specially engineered optical trench can be used to trap the energy
in the many modes which propagate within the fiber core
•
Keeping the light in the core, even in the most challenging bending
scenarios, significantly reduces the bend-induced attenuation
Trench acts like barrier
Energy is confined inside the fiber
Standard OM3/OM4 fiber versus
bend-insensitive OM3/OM4 fiber
9 High bandwidth OM3 and OM4
capability
9 Improved optical performance
9 Fully standards compliant; Compatible
with installed base
10
Macrobend loss
, 850 nm, 2 turns (dB)
9 Up to 10x better bend performance than
standard 50 µm fiber
9 May be spliced/ connectorized with
commercially available equipment
Bend Radius
Number of Turns
Max Induced Attn @ 850 nm
Proposed BI Fiber Spec @ 850 nm
1
Stand
ard OM
3/4 Fib
er
0.1
BendInsens
itive O
M3/4 F
iber
0.01
5
7
9
11 13 15 17 19 21 23 25
Bend Radius (mm)
Multimode Std
IEC 60793-2-10
Multimode Std
ITU – G.651.1
New Level of Bend
Performance
37.5 mm
15 mm
7.5 mm
100
2
2
0.5 dB
1 dB
0.05 dB
0.1 dB
0.2 dB
Typical Storage Area Network (SAN)
link includes > 30 bends
Bend-insensitive OM3 fiber
increases the spare margin
4
Max IL and Margin for 300 m 10G link
Total Insertion Loss (dB)
3
Protected “Headroom” or Spare
Operating Margin
2
1
0
Conventional
OM3 fiber
with bending
Loss
due to
bending
Chromatic
Dispersion
Improved Bend insensitive
OM3 fiber
Attenuation
with bend
Benefits of Bend-Insensitive OM3
Increase Spare
System Margin
Protect Link
Power Budget
Reliability is Key Concern for
System Designers
Cause of Downtime
•
•
Cables and connectors
accounted for 6% of downtime
Structured cabling can effect
43% of network downtime
Cost of Downtime
•
•
Cost of downtime varies based
on organizational size
Network degradations are
more difficult to trace
Bend-insensitive multimode fiber
enables better “box”
Substitute bendinsensitive multimode
fiber for conventional
50 µm fiber
B
A
Size of “box” with
conventional 50
µm fiber
•
Drawing To Scale
Loss of A = Loss of B
Size of “box” with
bend-insensitive
50 µm fiber
Key benefits:
•
•
•
•
•
Smaller, lighter, more compact cables, hardware and equipment designs
Reduced data center footprint
Better cooling/airflow; Reduced energy usage
Supports Green Data Center
Lower OPEX
Make connector side-pulls a
non-event
Standard 50 µm fiber
Bend-insensitive 50
µm fiber
Keep your network clear with
bend-insensitive multimode fiber
Standard 50 µm fiber
Bend-insensitive 50
µm fiber
Summary
• Multimode fiber remains the most cost-effective choice
for enterprise networks
• Bend-insensitive multimode fibers can help solve key
concerns of enterprise network operators
• OM4 fibers are now standardized and provide a path for
extended distances for next generation speeds
• Next generation standards will use OM3 and OM4 fibers
to provide low cost future-proof solutions for enterprise
networks