Fiber Trends, Applications and Architectures Rodney Casteel RCDD/NTS/OSP CommScope Technology Manager Chair TIA Fiber Optic LAN Section
Outline Bandwidth Demands and Industry Drivers Multimode Fiber Trends Fiber Applications Review of TIA 568 Architecture Standards Summary
Bandwidth Demands & Industry Drivers
Data at rest versus data in motion
Ethernet Switch Market Dell Oro Group Report (Aug 2009): Total Ethernet Switch market to rebound in 2010. Drivers: 10 GbE Data Centers Cisco s move into server market HP s expansion of Data Center portfolio IBM s entrance into Ethernet Switch market Ethernet switch market to begin rebounding in 2010. Lightwave Aug 4, 2009, with reference to report by Dell Oro Group telecom market research firm.
Datacom Transceiver Market Drivers / Notes: CIR Report (July 2009): By 2014 Datacom Tx market will grow 50% to $6.5B 10GbE market will reach $1.4B Fibre Channel market will reach $1.7B Processor speeds catching up with networking power of 10GbE. 10GbE equipped PC s expected in a few years. IEEE s 10GBASE SR fiber standard well suited for data centers, will remain dominant. IEEE s 10GBASE T copper standard hindered due to its huge power consumption. Datacom transceiver market to grow 50% by 2014. Lightwave Jul 22, 2009, with reference to report by CIR market research firm.
High Speed MM Transceivers Status and Trends 850 nm VCSEL captured the LAN market in <2 years of commercialization (now defacto std.) Reduced module size, power consumption and cost of gigabit optical components Many suppliers/low cost at the 1 Gbps level Example: 5 suppliers Advantageoptics.com Fiber 1000Base SX SFP transceiver average price $64.80 Copper 1000Base T SFP transceiver average price $157.00 1G serial fiber transceivers now less than half that of copper counterparts (SFP)
Market Driver Summary IP traffic growth continues at 30+% CAGR Switch market expected to grow in 2010 Transceiver market expected to grow 50% by 2014 Fiber transceivers quickly coming down in price as volumes increase Storage growth continuing at 35+% CAGR Faster speeds will be required in near future to deal with increased demands
Green Initiatives
Green Trends Faced with the on going debate of global climate change and rising energy costs and shortages, government agencies and private firms worldwide are examining ways to protect the environment. To address what is increasingly being perceived as a crisis, there is a growing global movement to implement more environmentally friendly computing and infrastructure
Think Long Term History has shown that the capabilities of the electronics used in our networks improve by a factor of 10 approximately every 5 to 7 years. Find your current data rate and select a cabling solution that will enable you to seamlessly transition to the next higher data rate to avoid the waste and expense of replacement In your planning, add a ZERO to your data rate every 5 to 7 years
Multimode Fiber Trends
OM4 Standardization
10 Reason For OM4 Standardization 1. OM4 supports all existing MM applications, making it a universal media 2. Maximizes cabling design flexibility for 850nm VCSEL applications from 1G to 100G 3. Extends reach to the longest distances of any MMF 4. Stretches power budgets to provide more operating margin 5. Supports the lowest cost transceivers for every application 6. Reduces energy costs by enhancing the utility of lowest power optics 7. Saves line card & switch chassis costs by extending the reach of optics much smaller than those for SM 8. Expands engineering space for higher rate application standards 9. Broadens usable spectrum for future 850nm WDM applications (e.g. possible 32GFC, 64GFC, 40GbE) 10. Reduces contamination sensitivity compared to SM connectivity Courtesy of Paul Kolesar, CommScope
OM4 Provides Superior Support for WDM OM4 supports broader wavelength spectrum at OM3 s 2000 MHz km bandwidth Courtesy of Paul Kolesar, CommScope
Size Matters: 100G MM and SM Transceivers MM 100G SNAP12 3-5 Watts SM Double XENPAK 16-18 Watts Diagram from 802.3 HSSG presentation cole_01_0107 (Finisar) Power dissipation from jewell_01_1106, pepeljgoski_01_0108, cole_01_0107, traverso_01_0308 SM ports consume line cards and chassis slots must faster than MM ports. Drives a significant SM cost disadvantage in Data Centers. Power dissipation will be much higher for SM given need for: 10G electrical to 25G optical lane rate conversion Thermo electric coolers to stabilize WDM wavelengths Edge emitter bias levels 5 7x 850nm VCSEL levels This is reflected in the heatsink size of the SM device Courtesy of Paul Kolesar, CommScope
100GbE Channel Cost Comparisons >7x >11x SM channel is >7x the cost of 10G SR 10xLAG huge market acceptance barrier Note: 10G SR channel costs would be ~1/3rd lower with SFP+ instead of X2 modules ~11x factor LAG can be deployed in >100m channels but necessitates the use of 10 ports per switch, and management of 10 separate channels XR offers >11x lower cost w/o increased port consumption or management!! Cost comparison does not account for module density impact on line card cost, which would amplify the illustrated XR cost factor and shrink the LAG cost factor XR module can be physically the same size as SR10 100m module no loss of density Courtesy of Paul Kolesar, CommScope
Multimode Roadmap Content courtesy of Corning, Draka, and OFS Sources cited include CRU, IEEE, Penwell, Cisco, IBM, Mathew Burroughs, Alan Flatman
Roadmap Multi-Mode Standards Data rate [Mb/s] 100000 10000 1000 100 10 1 10 M Ethernet Token Ring FDDI 100 M Fast Ethernet ATM LEDs 2 G 1 G FC GbE Fiber Channel Lasers 10 G 10 GbE 4 G FC 8 G FC 100 G GbE 40 G GbE 1985 1995 1998 2002 2005 2010 16 G FC OM1 / OM2 OM3 OM3 / OM4 Courtesy of Ryan Chappell, Draka Communications, BICSI Winter Conference 2010, Fiber Trends, Architectures and Cost Modeling
Historical Trend Multimode Fiber Types Related to Cost of Ports M.km Yearly demand per MMF type 2.5 2.0 /port 600 Cost of FC ports 1.5 1.0 0.5 0.0 1995 OM-1 ( 62.5 μm ) 1997 1999 OM-2 2001 2003 2005 OM-3 2007 KMI-2006 2009 500 400 300 200 100 0 2G 1G Source:Gartner 4G 8/10G 10GbE standard 1998 2000 2002 2004 2006 2008 2010 Courtesy of Ryan Chappell, Draka Communications, BICSI Winter Conference 2010, Fiber Trends, Architectures and Cost Modeling
Intel, Cisco, Yahoo data and predictions
Port shipments Intel's view of server forecast by Ethernet connection type (IEEE802.3HSSG tutorial) 1 GbE standard 10 GbE standard Fiber installation ahead of port shipments... Courtesy of Ryan Chappell, Draka Communications, BICSI Winter Conference 2010, Fiber Trends, Architectures and Cost Modeling
Port shipments High/Low speed switch port units ratio Cisco s view development switch connections (barbieri_01_0108) Courtesy of Ryan Chappell, Draka Communications, BICSI Winter Conference 2010, Fiber Trends, Architectures and Cost Modeling
MMF Applications / Data Center Links Total Optical Fibre Data Centre Links Total Data Centre Links by Media Type based on 40GBASĒ CR4/SR4 & 100GBASE - CR10/SR10 30 100GBE backbone links >125m 40GBE backbone links >125m 30 Total Links (millions) 25 20 15 10 100GBE backbone links <125m + 100GBE server links >10m 40GBE backbone links <125m + 40GBE server links >10m 10GBE backbone links <300m + 10GBE server links >100m 1GBE server links >100m Total Links (millions) 25 20 15 10 SMF Necessary MMF Possible Copper Possible 5 5 0 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 0 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 Source: Alan Flatman standards briefing 2009-11-09 One conclusion that could be drawn Multi-Mode will likely be around for the foreseeable future. Copper will likely also have a place. The copper/fiber decision will surely involve system cost, Green Data Center initiatives, etc. Second conclusion that could be drawn Single-mode will most likely not replace Multi-Mode in data centers Courtesy of Ryan Chappell, Draka Communications, BICSI Winter Conference 2010, Fiber Trends, Architectures and Cost Modeling
Analyst Reports CRU, Mathew Burroughs Global MMF demand Breakouts by type of fiber
2011 2012 2013 2014 North America Trend North America yearly percentage of MMF by fiber Type Data Source: CRU 2009 100% 90% 80% 70% 60% 50% 40% OM4 OM3 OM2 OM1 30% 20% 10% 0% 2004 2005 2006 2007 2008 2009 2010 Courtesy of Ryan Chappell, Draka Communications, BICSI Winter Conference 2010, Fiber Trends, Architectures and Cost Modeling
North America Trend Different Data Source North America Percentage by Fiber Type 2007-2009 Source - Mathew Burroughs Report 60% 50% 40% 30% 20% % OM1 % OM2 % OM3+OM4 10% 0% 2007 2008 2009 (Q1-Q3) Courtesy of Ryan Chappell, Draka Communications, BICSI Winter Conference 2010, Fiber Trends, Architectures and Cost Modeling
Global Growth of 10Gb/s Fiber Global Percentage of OM3+OM4 fiber worldwide 70% Data Source: CRU 2009 60% 50% 40% 30% 20% 10% 0% 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Courtesy of Ryan Chappell, Draka Communications, BICSI Winter Conference 2010, Fiber Trends, Architectures and Cost Modeling
Applications
40GBASE-SR4 and 100GBASE- SR10 Optical Lane Assignments
40GBASE-SR4 Interface Looking into the transceiver receptacle with the keyway on top Follow common themes of existing 4-lane variants TxTx TxTx RxRxRxRx Specify 4 transmitters on left side 4 receivers on right side 4 unused positions in center 40GBASE-SR4 Specify the MPO connector Male (pinned) in MDI receptacle Female (unpinned) on patch cord Requires same array cabling polarity as all existing 4-lane variants that transposes signals laterally P802.3ba lane striping protocol sorts lanes at receiver no need to assign lane numbers, but assignment may be desirable for diagnostic or other purposes Courtesy of Paul Kolesar, CommScope
100GBASE-SR10 Interfaces Three variants, all looking into the transceiver receptacle with the keyway on top Side-by- Side Ports Transmitter TxTx TxTxTx TxTxTx TxTx Receiver RxRxRxRxRxRxRxRxRxRx Vertically Stacked Ports Single Port Receiver RxRxRxRxRxRxRxRxRxRx Transmitter TxTx TxTxTx TxTxTx TxTx RxRxRxRxRxRxRxRxRxRx TxTx TxTxTx TxTxTx TxTx Specify Relative position of Rx and Tx Side-by-Side Ports follow IB Vertical and Single analogous 12-position rows with outer positions unused to mitigate alignment challenges Specify the MPO connector Pinned (male) in MDI receptacle Unpinned (female) on patch cord All variants can use the same 12-fiber array cabling polarity, required by HIPPI and IB, that transposes signals laterally, because lane striping sorts lanes. See examples in later slide. Courtesy of Paul Kolesar, CommScope
Fibre Channel Background T11.2 committee roadmap Completed 8GFC in 2008 Began 16GFC in 2009 8G highlights 16G Specifies traditional limiting receiver (150 m on OM3) Also equalized linear receiver (300 m on OM3) Products with limiting receivers available today Will use serial transmission (16G on single lane) Will specify distances on both OM3 and OM4 First ballot specified OM4 to 150 m
Trends for higher data rates Serial transmission will segue to inverse multiplexed transmission Space division multiplexing (parallel fibers) Wavelength division multiplexing (multiple colors) 802.3 adopted parallel fibers for 40G/100GbE 4f x 10G for 40GbE on OM3 10f x 10G for 100GbE on OM3 Exploring reach extensions on OM4 802.3 & FC already standardized 4λ transceiver for 10GbE Proposed: 2λ x 20G for 40GbE on MMF (λ near 850nm) Proposed: 4λ x 25G for 100GbE on MMF (λ near 850nm) OM4 would extend reach here also Courtesy of Paul Kolesar, CommScope
Solution Recommendation Given the trends towards: multi lane transmission (parallel & WDM) higher lane rates higher cost and power of SM solutions Relaxed transceiver specs leading to reduced supported distances Make Sure: You utilize a polarity scheme that allows straight forward migration to parallel optics without special polarity correcting components or a mixture of different patch cords You implement fiber counts in multiples of 12 at a minimum, but 24 is recommended You have the lowest guaranteed concatenated insertion loss available Use the highest grade multimode fiber you can afford, OM4 recommended
Review of ANSI/TIA/EIA-568-C Architectures
Structured Cabling System Architectures Evolving to Provide More Customer Options 1991: TIA 568 standard ratified Hierarchical star architecture Optimized for copper performance characteristics & limitations 100 meter horizontal cabling subsystem limit 1995: TSB 72 Centralized Fiber Optic Guidelines 2001: TIA 568 B.1 supports centralized fiber (FTTD) 2005: TIA 569 B & 568 B.1, Addendum 5 supports Telecom Enclosure (TE) 2009: TIA 568 B series and addendums were replaced with C series
Enterprise Networks Traditional Hierarchal Star Main cross connect in equipment room with fiber backbones to remote Telecommunications Rooms/Closets (TR/TC) Require floor space for the passive Horizontal Cross Connect Copper Horizontal. Centralized Fiber Implementation of Fiber To The Desk (FFTD) Main Cross Connect in Equipment room with fiber backbones that end at user workstations (fiber horizontal) TR/TC only functions as passive optical interconnect/patch Fiber To The Enclosure (FTTE) & Zone Main cross connect in equipment room with fiber back bones to remote mini TR (typically an enclosure) Copper Horizontal ISO/IEC 11801 Data Distributor
Traditional Hierarchical Star Topology Copper Opportunity: TR Patching Horiz. Cabling (long) Work Area Cabling Fiber Opportunity: ER Patching TR Patching Backbone Cabling
Pros: Enterprise Architectures: Hierarchical Star Pros/Cons Historical Acceptance - Industry Standard Large number of users serviced by a single TR allowing for per floor local patching and administration POE capability via switch and/or midspan Easy to deliver dedicated/back-up power to TR Easy to deploy hybrid solutions, i.e. bus, star, ring etc. Workgroup switches/devices are in central location Easy cable installation being able to terminate and pull multiple runs simultaneously from a central point Easier to secure equipment in a TR than in distributed locations
Cons: Enterprise Architectures: Hierarchical Star Pros/Cons Infrastructure Life - short compared to FTTD or FTTE Scalability - Incremental build (new TRs) time consuming & expensive Floor space req's - cost of floor space is high Large TRs affect more users during outages 24/7 Heating & Cooling Requirements May not easily migrate to converged networks in future 90m restriction, limits architectural flexibility Potentially yields inefficient use of switch ports May lead to inefficient port utilization
Centralized Fiber Topology (FTTD) Copper Opportunity: ER Patching Fiber Opportunity: ER Patching TR Interconnect Backbone Cabling Work Area Cords
Components of FTTD FTTD According to TIA 568C.1 fiber can run directly from the main equipment room to the desk top for all runs 90 meters or less, but anything over 90 meters requires the fiber to pass through a TR and breakout to a patch panel or splice point. With a FTTD architecture all the electronics are centralized within the main equipment room allowing for better port utilization. At the desktop the fiber can route into a fiber NIC or go through media converters. Typically, small fiber count cables of 2 4 fibers are home run allowing for better pathway/space utilization In the main equipment room, copper ports are utilized between the core switch and workgroup switches, which saves money on equipment cost.
Enterprise Architectures: Centralized Fiber Pros/Cons Pros: Highest Potential Non-Blocking Bandwidth 100% of SCS (outside of ER) is EMI indifferent Security - easy to administer - no switch gear outside of ER Longest life infrastructure - media and connectivity Longer horizontal potential (>100m) Minimal (or no) TR req'd Security only one location to secure Smaller/lighter cabling all the way to the desk - compared to copper Most efficient use of switch ports Centralized power and back-up
Cons: Enterprise Architectures: Centralized Fiber Pros/Cons Cost - initial installed cost may be the highest of any architecture, depending on number of users per floor and media converters vs. NICs Does not scale well - Infrastructure changes more difficult Familiarity & Acceptance - Fear of Fiber! May not migrate well to a converged network Cable congestion in equipment room could become an issue Does not support POE, but can support other power options
FTTE (Fiber To The Enclosure) Topology Copper Opportunity: TR Patching Horiz. Cabling (short) Work Area Cords Fiber Opportunity: ER Patching TE Patching Backbone Cabling
FTTE/Zone Components of FTTE Unlike FTTD where the TIA requires a pull through for distances greater than 90 meters, the FTTE topology is allowed to go the 300 meter distance to the TE. The telecom enclosure (TE) becomes the horizontal cross connect. The TE allows for the same type of application deployment that a traditional TR allows, but on a much smaller scale, i.e. POE, Wireless, Copper, Fiber etc. The actual design can allow for a totally non blocking network, as well as getting the larger bandwidth media closer to the end user. The FTTE design will typically be the lowest cost topology. The savings come from the real estate utilization, lower cost work group switches, port utilization and being able to utilize copper NICs instead of fiber NICs.
Typical Hierarchical Star with UTP
ER Typical FTTE/Zone with Fiber and UTP
Pros: Enterprise Architectures: FTTE/Zone Pros/Cons Quick Deployment with lowest cost Flexible - for integration of new tech. (BAS, POE, wireless, etc.) Scaleable - with minimal cost & disruption Easy MACs - intuitive to manage Non-permanent with no floor space req'd Network disruption (maintenance) - Reduce # users affected by downtime Simple cable install in horizontal - extends backbone distance >100m Opportunity to route fiber closer to the desk - Future ready Is the recommended architecture for converged networks in the ANSI/TIA/EIA-862 Building Automation Systems Standard Easily accommodate power users Greener and more sustainable architecture than Hierarchical Star
Cons: Enterprise Architectures: FTTE/Zone Pros/Cons Switches in enclosure (possible security concerns) Possible heat dissipation concerns in work areas Limitations on user count serviced by individual enclosure Each enclosure requires dedicated power if active Distributed electronics may be more difficult to manage compared to a centralized approach If enclosure is ceiling mounted, then MACs would require a ladder
FTTE Solutions Typical TE Enclosures
Summary
Summary Bandwidth requirements and data rates will continue to increase leading to the demand for higher grade multimode fiber and larger fiber counts New cabling architectures may offer high performance, cost effective alternatives As the number of optical ports continue to increase, the cost of optical equipment will continue to quickly drop making fiber architectures more favorable Green initiatives will continue to promote architectures which consolidate and utilize the least amount of natural resources and consume the least amount of power No option is perfect for every environment so consider all factors and choose based on individual requirements FOLS Cost Model is a good resource for evaluating architecture options (www.fols.org)
Thank You Questions? Rodney Casteel RCDD/NTS/OSP CommScope Technology Manager Chair TIA Fiber Optic LAN Section