CWDM and OEO Transport Architectures

Transcription

1 CWDM and OEO Transport Architectures Sajjad Hussain (Marconi) and Ross Halgren (RBN) Abstract This paper outlines the features and service benefits of integrated CWDM and OEO metro transport architectures. Following an overview of the xwdm product and network evolution and the ITU-T G.695 standard for CWDM architectures, examples of OEO transport platforms such as Multi-Service Provisioning Platforms (MSPPs) and Regenerative CWDM (R-CWDM) transport platforms are presented. The need for 10G services is discussed with reference to current and future OEO network implementations. Expansion of existing 2.5G R-CWDM platforms to accommodate future 10G xwdm OEO upgrades is foreshadowed using new, recently announced photonic and electronic technologies. 1. xwdm Evolution 1.1. Network Evolution In the late 90s, DWDM was introduced to lower the cost of adding capacity to long haul transport systems. The cost of laying new fibre and deployment of regenerator systems over long distances readily justified the DWDM alternative. 10G systems with 40 wavelengths were commonly deployed and systems with up to 160 wavelengths were touted. TDM grooming into 2.5G and 10G channels occurred at the edges of the long-haul and eventually, all-optical switching and optical performance monitoring were provided at major nodes to increase the flexibility of what was previously a dumb fibre pair-gain solution. Whilst it was known that metro fibre was more abundant than long-haul fibre, there were pockets of fibre exhaust that required an xwdm solution and reusing a long-haul DWDM product was the first approach. However, it was soon realized that metro DWDM networks differed from long haul networks by: the equipment size; power; cost; the number of wavelengths required and the number of different services and protocols to be supported on each wavelength. In the absence of an ITU standard such as Generic Framing Procedure (GFP) for multiplexing multiple different protocols into a 2.5G or 10G TDM stream, the simplest method of transporting multiple services was to allocate a wavelength for each service and transport the service via its native protocol format and rate Product Evolution Due to their long-haul heritage (some would say baggage ), first generation metro DWDM products had multiple variants of line and client interface cards as well as different OADM filter options. For example, there was a different line interface card for each DWDM wavelength and a different client interface card for each protocol (eg, Gigabit Ethernet, Fast Ethernet, Fibre Channel, ESCON, DVB- ASI, STM-n) and for each client wavelength (850nm, 1310nm and 1550nm). Additionally, there were different OADM filters for each band of four DWDM wavelengths, or a different filter per wavelength in some extreme cases. This resulted in an ordering and sparing nightmare for both the carriers and the suppliers. In contrast to long haul networks, metro networks by their very nature require greater reconfiguration to accommodate customer churn. This resulted in truck rolls to remote OADM nodes to change DWDM filters and DWDM line cards. All-optical metro DWDM rings included EDFA amplifiers to overcome accumulated filter losses. The need for per-wavelength optical power balancing when services were added or removed added further product complexity and cost. In 2001, CWDM technology was introduced to lower the cost of metro network deployments. As shown in Figure 1, CWDM used wider (20nm) wavelength spacing and this enabled the use of low cost uncooled lasers and low cost thin film optical filters. Suppliers who designed new metro CWDM products from the ground-up achieved the lowest cost-points. Incumbent suppliers who simply swapped-out DWDM line cards with CWDM line cards achieved the least cost erosion and the greatest margin erosion. CWDM and OEO Transport Architectures - IEE Paper.doc Page 1

2 Figure 1 ITU-T G CWDM Wavelength Grid For both the long haul and the high-end metro applications, tunable DWDM lasers were also introduced in 2001 to reduce the number of line card variants and sparing problems. For both DWDM and CWDM metro products, programmable multi-protocol, multi-rate client interface cards with pluggable optics in the form of GBICs were next introduced to overcome the client interface variants and sparing problems. This was quickly followed by Small Form Factor Pluggable (SFP) transceivers which are now widely adopted on all forms of equipment requiring optical interfaces not just xwdm equipment. SFPs are available for bit rates up to 2.7 Gbit/s and more recently 4 Gbit/s (for 4G Fibre Channel). They include a serial I 2 C digital diagnostics interface for monitoring device and signal parameters. On the line side, tunable CWDM lasers were more challenging so CWDM SFPs supporting rates up to 2.7 Gbit/s helped reduce the cost of multiple CWDM line card variants and associated sparing problems. This has been quickly followed by the recent introduction of DWDM SFPs for rates up to 2.7 Gbit/s and with 100GHz spacing. A DWDM FSP MSA forum [1] has been recently formed to develop an industry standard for DWDM SFPs. However, the problem of multiple OADM filter variants and the need for truck rolls to change them has not been solved by the introduction of CWDM technology. To overcome the distance limitations caused by accumulated OADM filter losses in all-optical CWDM networks, Semiconductor Optical Amplifiers (SOAs) were introduced in However, SOAs have not proven to be successful to-date due to: the small number of CWDM wavelengths supported (typically only four); optical power saturation problems; and the reintroduction of optical power balancing complexities and costs reminiscent of the high-cost, all-optical metro DWDM networks. For 10G bit rates, a variety of pluggable grey optics transceivers (supporting 850nm, 1310nm and 1550nm wavelengths) have been introduced, however, the XFP variant is increasing in popularity due to its small size, power, cost and simple serial interface. 10G XFP pluggable coloured optics are expected in the near future in either or both CWDM (20nm) and DWDM (100/200GHz) wavelength spacing. While SFPs include an increasing number of features such as multi-protocol support and digital diagnostics, XFPs epitomize the new generation of smart pluggable optics which include: FEC optimization with eye shape control, sampling position and decision threshold control for maximum BER; DWDM wavelength tuning; integrated multi-rate CDRs ( Gbit/s) and external clock synchronization; temperature monitoring and Thermo-Electric Cooler (TEC) control; high speed serial data and management interfaces; and EEPR for storage of configuration data, CLEI codes etc. CWDM and OEO Transport Architectures - IEE Paper.doc Page 2

3 Given the wider range of features enabled by XFPs for DWDM interfaces, it is not yet clear whether there will be demand for CWDM XFPs for 10G applications. There may not be sufficient costdifferential between CWDM and DWDM XFPs to justify two different variants, especially given that CWDM was originally intended to use simple, un-cooled lasers for low cost metro applications. The above active component evolution has not yet solved the problem of multiple OADM filter variants for DWDM and CWDM systems. Tunable filters are likely to offer a solution to this problem for future DWDM systems. Tunable filters may also prove to be a lower cost alternative to MEMS optical switches for remotely reconfigurable OADM nodes (thus avoiding truck rolls), however, there is again no low cost technology for implementing remotely reconfigurable, all-optical CWDM networks. 2. ITU-T G.695 Standard CWDM Architectures Whilst the ITU-T G defined the CWDM wavelengths and optical filter bandwidths, a new standard, G.695 was required to standardize on the CWDM links, network element architectures and reference point parameter values. A set of application codes have been specified that represent various CWDM network implementations. The following sections describe the main architecture options Black Box Uni-Directional Architecture Most readers will be familiar with the Black Box architecture option illustrated in Figure 2. All long haul DWDM products and first generation CWDM products are based on this approach. Each network element (such as a terminal mux) includes a number N of CWDM lasers connected to a CWDM optical multiplexer (filter). The output fibre interface of the CWDM Optical Multiplexer defines an accessible transmit reference point MPI-S M which is now standardised by G.695. Similarly, there is a receive reference point MPI-R M at the input to a remote CWDM Optical Demultiplexer (filter) which in turn is connected to N receivers. For a bi-directional CWDM link, the network elements shown in Figure 2 will be repeated in the reverse direction on a 2 nd fibre, generally using the same wavelengths. Thus two fibres are required for bi-directional transmission channels. Tx λ 1 Rx λ 1 Tx λ 2 Rx λ 2 MPI-S M MPI-R M Tx λ N Rx λ N CWDM Network Element Figure 2 CWDM Network Element Black Box - Uni-Directional Architecture Figure 2 only depicts the CWDM line interfaces to a product. If this product happens to be a switch or router for example, then such interfaces may be deemed client interfaces with respect to another transport product that is required to interface to it. If instead, this product happens to be a multichannel media converter and CWDM multiplexer, then again, only the CWDM line interfaces are shown. In this case, the client interfaces (not shown) may be electrical or optical, and if optical, the client interfaces may be 850nm multimode, 1310nm singlemode or 1550nm singlemode for example (referred to as grey optics). Such a multi-channel media converter and CWDM multiplexer may CWDM and OEO Transport Architectures - IEE Paper.doc Page 3

4 include vendor-defined features such as SFP client interfaces (for low up-front cost and sparing), 3R regeneration and client signal performance monitoring (in each direction). For end-end management, an Optical Supervisory Channel (OSC) may be carried over a CWDM wavelength, however, the OSC information and protocol have not been standardised under G Black Link - Uni-Directional Architecture The Uni-Directional Black Link architecture illustrated in Figure 3 was standardised to accommodate multi-vendor CWDM networking possibilities especially those enabled by the emergence of coloured pluggable optics based on the GBIC and SFP formats. As shown in Figure 3, standard reference points are defined at each CWDM laser output (S S ) and receiver input (R S ) as well as the Optical Multiplexer output (RP S ) and Optical Demultiplexer input (RP R ). In this case, each laser and receiver may be associated with different equipment vendor s products and the external Optical Mux and Demux filters may also be supplied by other vendors. CWDM Link Tx λ 1 S S R S Rx λ 1 Tx λ 2 Rx λ 2 S S RP S RP R R S Tx λ N Pair of CWDM Network Elements Rx λ N Figure 3 Black Link Uni-Directional Architecture Bi-directional links will replicate the equipment shown in Figure 3 for the reverse direction using a 2 nd fibre and generally using the same wavelengths. In this case, the CWDM link is a dumb fibre pairgain system. There are no 3 rd party value-added features for a carrier to use - such as multi-protocol performance monitoring of each channel or service demarcation point. Features such as OSC implementation are left to the individual equipment vendors hence total interoperability between different systems remains a challenge. In addition, it is difficult for a carrier to offer Service Level Agreements (SLAs) based on the Black Link architecture Black Box Bi-Directional Architecture Figure 4 shows the Black Box architecture for Bi-Directional Single Fibre Working (SFW) applications. The benefits of SFW over two-fibre links are reduced fibre requirements, reduced up-front fibre leasing costs and simpler installation (since can t get the Tx & Rx fibres around the wrong way). The cost benefits are accentuated in the case of protected links, for which the uni-directional per fibre approach requires four fibres per protected link and the bi-directional per fibre approach requires only two fibres per protected link. The bi-directional approach thus potentially releases two fibres to support another service network (such as a 10 Gigabit Metro Ethernet ring or a STM-64 NG-SDH ring). CWDM and OEO Transport Architectures - IEE Paper.doc Page 4

5 CWDM Network Element CWDM Network Element Tx λ 1 Rx λ 1 Pluggable Optics or Integrated Laser and Receiver Arrays Rx λ 2 MPI-S M MPI-R M Tx λ 2 MPI-R M MPI-S M Tx λ N-1 Rx λ N-1 Pluggable Optics or Integrated Laser and Receiver Arrays Rx λ N Tx λ N Note that λ K (K = 1.. N) does not correspond to any particular wavelength or order Figure 4 Black Box Bi-directional Architecture Figure 4 also more conveniently illustrates two implementation options for a Black Box CWDM line interface. Pluggable Coloured Optics One Black Box implementation that is increasingly being used in new CWDM products is pluggable coloured optics (GBIC or SFP). This reduces the up-front cost to some extent, since the pluggable CWDM transceivers only need to be purchased when the extra CWDM channels are required. This approach also eases the sparing costs compared to dedicated, wavelength-specific line interfaces. However, the potential savings of this approach are diluted due to the extra cost of manufacturing a pluggable transceiver and associated motherboard housing (bezel), and the middle-man margins associated with such value-added products. The extra features offered by XFPs for future 10G applications will only accentuate this value-added cost. The bottom line is that the carrier is unwilling to pay more for this feature, so the total product margin is split between the end-product supplier and the CWDM SFP/GBIC/XFP transceiver supplier. Integrated Array Optics Another Black Box implementation that draws on the inherent low cost of CWDM components is the Array Optics approach. The premise behind this approach is that lasers, drivers, receivers and CWDM filters in raw form cost considerably less and consume less space when integrated than separately packaged SFP/GBIC modules and CWDM Mux/Demux filter modules. Comparison of Implementation Options CWDM SFPs may comprise a laser TOSA (Transmitter Optical Sub-Assembly) to which is added a laser driver, PIN/APD and TIA receiver, optical signal monitoring, EEPR and I 2 C digital diagnostics interface. These components are packaged in a SFP module with electrical edge and optical LC connectors. To this is added the board-mounted bezel into which the SFP module plugs. An 8- wavelength bi-directional Black Box interface would require four such SFP modules and bezels, a separate CWDM Mux/Demux filter module and four duplex optical interconnect cables fitted with LC connectors. In contrast, an array-optics implementation may integrate four laser drivers, four TOSAs - each of a different wavelength, four APD/TIA receivers and CWDM Mux/Demux filters into a single 8-wavelength line interface module. Using pigtailed CWDM laser and receiver components spliced to CWDM Mux/Demux filter components eliminates LC connector costs, optical jumper cable costs and connector db losses. The latter feature reduces the CWDM Mux/Demux insertion losses and thus longer transmission distances are possible compared with pluggable coloured optics based CWDM implementations. CWDM and OEO Transport Architectures - IEE Paper.doc Page 5

6 For a fully configured, 8-wavelength bi-directional Black Box implementation, the cost to the endproduct supplier of the integrated array-optics approach is less than the cost of four pluggable coloured optics modules with a separate CWDM Mux/Demux filter module and associated optical interconnect cables. The cost difference is attributed to the elimination of the value-added costs introduced by the SFP and filter module suppliers. Of course, there are no free lunches, since the cost difference can only be achieved through custom design and development which bares a one-off R&D cost. For a partially configured bi-directional Black Box configuration, there is a cost per channel break-even point between the two implementations. Typically this corresponds to approximately two bi-directional channels, after which the integrated array optics approach is cheaper. Given that few practical CWDM links will have less than two bi-directional channels, the integrated array optics approach is the superior alternative for simple optical pair-gain applications. In additional to reducing the cost of multi-channel CWDM links, the array optics approach overcomes the parts variant and sparing problems normally associated with xwdm systems problems which tunable DWDM lasers and pluggable coloured optics were originally developed to alleviate. This results in operational cost savings for the carrier or service provider, since there are no longer any wavelength-specific modules to purchase instead, there is just an East module and a West module. As will be revealed later, the integrated array optics approach also facilitates flexible service network solutions that offers much more to the carrier than a simple optical pair gain solution to a fibre exhaust problem Black Link Bi-Directional Architecture Figure 5 illustrates the G.695 Black Link Bi-Directional (SFW) architecture. This approach is suitable for multi-vendor transport implementations using SFP, GBIC or XFP based pluggable coloured optics fitted to each vendor s equipment (switch, router, server, etc). In this case, the CWDM Mux/Demux filters will manage the bi-directional wavelength multiplexing function; consequently bi-directional pluggable transceivers are not required (often referred to as BiDi transceivers). Standard SFP/GBIC/XFP CWDM transceivers may be used since their PIN/APD detectors are broadband ie, capable of receiving all ITU-T G CWDM wavelengths. CWDM Link SFP / GBIC Tx λ 1 Rx λ 2 S S R S RP S RP R R S S S Rx λ 1 Tx λ 2 SFP / GBIC SFP / GBIC Tx λ N-1 Rx λ N S S R S RP R RP S Pair of CWDM Network Elements R S S S Rx λ N-1 Tx λ N SFP / GBIC Note that λ K (K = 1.. N) does not correspond to any particular wavelength or order Figure 5 Black Link Bi-directional Architecture As for the Black Link Uni-Directional architecture, the drawbacks with this approach are the lack of management features needed to support carrier environments, including the lack of separate, multiprotocol performance monitoring to meet SLA needs, and the lack of a separate OSC for independent remote configuration and supervisory functions. The Black Link approaches are more suited to enterprise customers having their own dark-fibre networks. CWDM and OEO Transport Architectures - IEE Paper.doc Page 6

7 2.5. Hybrid Black Link and Black Box Architectures CWDM networks must be able to support both old and new transport equipment and client equipment interfaces. Consequently, there are a number of applications where a hybrid Black Link and Black Box architecture is required. One such application is illustrated in Figure 6, for which new generation equipment with pluggable coloured optics interfaces on the left must interface with old generation equipment with grey optics interfaces on the right. In this example, a media converter function is required on the right to convert from the grey to the coloured interfaces. Such a media converter includes an Optical-Electrical-Optical (OEO) stage. This OEO stage may additionally include features such as 3R regeneration using multi-protocol, multi-rate Clock Data Recovery (CDR) devices and multi-protocol performance monitoring in accordance with ITU, Telcordia and IEEE standards. Coloured or Integrated Laser and Receiver Arrays Grey Service 1 Tx λ 1 Rx λ 2 Rx λ 1 Tx λ 2 Trib Tx Trib Rx Service 1 CWDM Electrical 850 / 1310 nm Service N Tx λ N-1 Rx λ N New Generation GFP-Multiplexers STS/VC Switches Routers & Servers CWDM Network Elements Rx λ N-1 Tx λ N Trib Tx Trib Rx Note 1: λ K (K = 1.. N) does not correspond to any particular wavelength or order Note 2: This architecture option is derived from the G.695 standard it is not part of the standard. Service N Old Generation Multiplexers STS/VC Switches Routers & Servers Figure 6 Hybrid Black Link & Black Box - Bi-directional Architecture 3. OEO Transport Platforms 3.1. Background Several factors contributed to the deployment of all-optical DWDM in the mid 90 s. The convergence of 2 nd generation photonics technology and the envisaged bandwidth demands fuelled by the Internet and the application hosting services market contributed to massive investment seen in most of the service provider core networks. Prior to this technical and market revolution, 1 st generation optical networks dating back to the early 1980s were based on OEO architectures. Examples included PDH carrier networks, FDDI enterprise networks, ATM networks, SONET/SDH networks and DQDB metropolitan networks. Such OEO networks continue to proliferate, with Next Generation SONET/SDH networks, Metro Ethernet networks and Resilient Packet Ring (RPR) networks being prime examples. Additionally, IP and MPLS networks rely on OEO architectures for their existence. The reason for the success of OEO transport architectures in carrier networks can be understood in terms of where carriers make money and where they don t. Carriers learnt in the late 80s that there was little value add - and therefore revenue - in selling dark fibre to end customers. More recently, this principle can also be applied to carriers offering dark wavelengths (by selling Black Link networks for example). By contrast, Carriers networks add value and can, therefore, attract a premium for selling SLA driven services and securely managing customer information. However, the intense competition to provide bandwidth at almost any price to the end-user has reduced the inherent value of raw circuit pricing levels to a commodity level. Operators often talk about moving from specialized circuit switchbased platforms to softswitch-based platforms and it is the value added functions involving multiprotocol circuit and packet switching, data processing, performance monitoring and information management where carriers make their money. Such functions are bundled into a range of services and priced accordingly. These functions are still implemented in the electrical domain rather the optical domain - especially as the granularity of information and bandwidth gets smaller and smaller. CWDM and OEO Transport Architectures - IEE Paper.doc Page 7

8 3.2. Multi-Service Provisioning Platforms (MSPPs) MSPPs are next generation SONET/SDH ADMs and Edge Switches which provision, monitor and manage multiple services of various types. They support both circuit and packet switched services (ranging from at least VT-1.5/VC-12 circuit granularity up to 10 Gigabit Ethernet packet services). To support such a wide range of services and bandwidths, MSPPs often include circuit and packet backplanes, STS/VC circuit switching cards and packet switching cards (supporting Layer 2 and/or Layer 3 switching and routing functions) and a GMPLS management control plane. New generation MSPPs use the ITU Standard G.7041 / Y.1303 Generic Framing Procedure (GFP) for multiplexing multiple different protocols into a SONET/SDH frame for transport between MSPP nodes and for STS/VC or packet switching between line interfaces and tributary interfaces within MSPP nodes. Protocols supported by GFP include for example, Gigabit Ethernet, Fast Ethernet, ESCON, 1 Gig Fibre Channel, 2 Gig Fibre Channel and DVB-ASI. Traditional services that use T1/E1 and T3/E3 are mapped, switched and transported in the usual way over VT-1.5/VC-12 and STS-1/VC-3 TDM channels respectively. Performance monitoring is provided for all protocols and channels to support Service Level Agreements. Simple MSPP ADMs will generally have two line interface cards (east and west), at least one switch card (or two for redundancy) and multiple tributary interface cards, each supporting multiple tributary ports. Additional MSPP line interface cards can be installed to support mesh network and crossconnect switch applications. In some cases, the line interfaces may be integrated with the switch cards to save space and cost. New generation MSPPs employ grey and coloured SFP optics as well as standard electrical connectors on the tributary ports. Diagnostic loop-backs and optical signal monitoring are provided on all optical line and tributary interfaces. MSPPs currently support both 2.5G and 10G line interface options (with or without OTN / FEC options). Grey or coloured SFPs with CWDM and DWDM options are available for 2.5G line interfaces. Future 10G line interfaces are expected to support grey or coloured XFP optics. For single channel, Single Fibre Working (SFW) link configurations, 1310nm and 1550nm SFPs are used at each end of a link (with an external WDM filter to separate the two directions of transmission). The CWDM and DWDM SFP options are generally used in a Black Link configuration where multiple MSPP or other carrier platforms must share the same fibre link due to fibre exhaust problems. In this case, the Black Link approach is acceptable to the carrier since the MSPP based OEO platforms include the necessary 3R regeneration, performance monitoring and DCC (or OTN) management channels that are expected in a carrier environment. An example of a simple MSPP in a SFW configuration is illustrated in Figure 7. Grey Tributary Interfaces Line Interface 1 Trib Rx Trib Tx 850 nm 1310 nm Trib Rx Trib Tx Line Interface 2 Multi-Protocol GFP Multiplexing 3R Regeneration Telcordia / ITU / IEEE Performance Monitoring Cross-Connect STS/VC Switch Diagnostic Loop-backs Remote Management via DCC or OTN SFW SONET/SDH/xWDM Transport O Rx 1550 Tx 1310 TDM TDM Tx 1550 Rx 1310 O SFW SONET/SDH/xWDM Transport Coloured Fixed or Electrical Processing Switching and Multiplexing Coloured Fixed or Figure 7 Next Generation SDH - MSPP ADM Platform CWDM and OEO Transport Architectures - IEE Paper.doc Page 8

9 An example of a MSPP product that is currently being deployed in carrier networks throughout the world is the Marconi S Regenerative CWDM (R-CWDM) All-optical CWDM and DWDM networks often employ Optical Add/Drop Multiplexers (OADMs) to connect specific pairs of tributary (client) interfaces in a fault-tolerant ring configuration. In the case of metro network applications, remote reconfiguration of connections is generally not available due to the relative high cost of optical switching and tunable lasers. Consequently, dedicated filters and xwdm transceivers are installed and hard-wired. Changes in connectivity as customers needs change requires a truck roll to remote sites to change filters and xwdm transceivers. All-optical CWDM and DWDM networks are characterized (and hindered) by accumulated optical losses and dispersion. For multi-node rings with OADMs that employ pluggable xwdm optics, the endend optical loss rapidly accumulates due to OADM filter insertion losses, filter-to-transceiver patch cable connector losses and the sum of all fibre link losses. Applications requiring broadcast xwdm channels incur even greater insertion loss due to optical power splitters. For 2.5G CWDM channels, accumulated dispersion at 1610nm can be a major problem which results in excessive dispersion power penalty. The use of SOAs with Variable Optical Attenuators (VOAs) for power balancing in CWDM OADM networks is a costly solution to the accumulated loss problem and is limited in effectiveness due to SOA saturation and noise figure problems and their inability to overcome accumulated dispersion at 1610nm. DWDM OADM networks can use EDFAs and VOAs as a better alternative to SOAs to overcome the accumulated loss problems, but such solutions are still relatively expensive. DWDM networks employing C-band wavelengths (shown in Figure 1) do however, have the benefits of lower fibre loss and less dispersion than CWDM networks, and they are able to support 10G channels. The above accumulated loss and dispersion problems of all-optical metro CWDM networks severely limit the network size when four or more OADMs are installed. Whilst individual point-to-point links of up to 80km at 2.5G rates are possible, the effect of accumulate loss and dispersion can limit the network size (eg, ring perimeter) to less than 40km in a multi-node OADM network. To overcome this limitation, all-optical xwdm suppliers now provide optional 3R regenerator cards which can receive, regenerate and retransmit a selected wavelength that is suffering from excessive loss and dispersion. These cards consume precious space and add extra cost to each OADM node. For any pair of OADM nodes that are more than 60km apart, it is highly likely that both nodes will require a 3R regenerator card for each wavelength transported. All the above problems can be overcome by using Regenerative CWDM (R-CWDM) OADMs [2] based on a back-to-back G.695 Black Box architecture with a multi-wavelength, integrated array optics implementation on each of the west and east CWDM line interfaces. As stated previously in section 2.3, committing from the outset to an integrated array optics implementation can actually reduce the cost of CWDM optics compared to alternative pluggable coloured optics implementations. In doing so, the path is clear to add additional OEO features and benefits at minimal extra cost, resulting in a platform that offers much more for the carrier than a simple fibre pair-gain solution to a fibre exhaust problem. As illustrated in Figure 8, these OEO features include for example, multi-protocol 3R regeneration of all wavelengths, thus overcoming the network size limitations of all-optical metro networks. Another feature that can be added includes protocol agnostic electrical cross connect switching (using a nonblocking M x M space switch). This provides the benefit of remote reconfiguration thus avoiding the cost of truck rolls and CWDM filter and transceiver changes when customers connectivity needs change. Such switches also support lossless drop-and-continue for broadcast applications and hairpinning for local switching applications. Multi-protocol performance monitoring on all tributary interfaces can be added, thus enabling SLAs between the carrier and its enterprise customers. Diagnostic loop-backs on all tributary interfaces and all wavelengths on the line interfaces provide the benefit of remote fault localization (and reduced operational costs) and increased network availability. All these features can then be managed via an Optical Supervisory Channel (OSC). CWDM and OEO Transport Architectures - IEE Paper.doc Page 9

10 Comparing the two OEO platforms presented (MSPP and R-CWDM), it is apparent that a Regenerative CWDM platform offers many of the features of a mini MSPP. Following this likeness, adding GFP multiplexing capabilities to the R-CWDM platform tributary interfaces adds the benefit of improved wavelength utilisation and increases the number of services that can be supported on a single fibre strand [3]. Grey Tributary Interfaces Line Interface 1 Trib Rx Trib Tx 850 nm 1310 nm Trib Rx Trib Tx Optional multi-protocol GFP Multiplexing for improved capacity utilisation Line Interface 2 Rx λ 1 Tx λ 2 Multi-Protocol 3R Regeneration Tx λ 1 Rx λ 2 SFW CWDM Transport Rx λ N-1 Tx λ N Telcordia / ITU / IEEE Performance Monitoring Protocol Agnostic Cross-Connect Space Switch Diagnostic Loop-backs Tx λ N-1 Rx λ N SFW CWDM Transport Rx λ OSC Tx λ OSC Remote Management via OSC Channel Tx λ OSC Rx λ OSC Integrated Array Optics Electrical Processing and Switching Integrated Array Optics Figure 8 Regenerative-CWDM OADM with Optional GFP Multiplexing Using an 8-wavelength CWDM Black Box Bi-Directional implementation, the above architecture provides an extremely flexible and low cost R-CWDM platform having up to 10 Gbit/s bi-directional capacity per fibre strand. By using simple, low cost, protocol agnostic media conversion techniques initially, the platform can be as small as a 1RU x 19 x 12 shelf and thus can be deployed in a standard carrier or enterprise network environment. The addition of GFP functionality to each R-CWDM node using a plug-in or stack-on modular implementation then enables a pay-as-you-grow approach to providing additional lower speed services. A typical Metro CWDM ring using the Regenerative CWDM platform approach is illustrated in Figure 9. As can be seen from this illustration, a wider range of services and interface options are supported and in contrast to similar multi-node, all-optical CWDM networks (carrying 2.5G channels), the R-CWDM based metro network size is no longer limited to 80km. Due to its flexible, modular architecture, the power of the R-CWDM platform does not compromise the cost of simple unprotected point-point network applications. In this case, only one CWDM line interface is required and the platform reduces to that outlined in section 2.3 for the simple Black Box Bi-Directional array-optics implementation. An example of an 8-wavelength R-CWDM product that has been deployed in carrier networks throughout the world is the RBNi GigaEdge CWDM and OEO Transport Architectures - IEE Paper.doc Page 10

11 Figure 9 Regenerative-CWDM Metro Ring 4. Service Networks In the 1990s OEO platforms such as DQDB multiplexers and ATM switches aimed to support multiple protocols (Ethernet, Frame Relay, X.25 etc) by adding Adaption Layers to convert from the native protocol to the new high speed (DQDB / ATM) protocol. The new high speed protocols could be either sent in their native format or could be mapped into SONET/SDH virtual tributaries or containers. For each client protocol that existed or was later developed, a new adaption layer had to be specified. Each adaption process added delay and cost to the transport process. More recently, similar aspirations were sought by using Gigabit Ethernet, IP or MPLS as a high-speed transport protocol to which all other protocols must be converted. As technology moves on, new protocols are continually being developed to meet the needs of specific markets and services (eg, 1G, 2G, 4G and 10G Fibre Channel for SAN applications). Inevitably, new higher speed protocols will be developed and optimized for various new applications and services. The introduction of transparent xwdm and more recently transparent GFP mapping has changed the mindset of the telecommunications industry from one of conversion to a single super-protocol, to one of support for multiple protocols each optimized for a particular service. The result is a service network approach to the design of the carrier s transport and switching infra-structure. The previously described MSPP and R-CWDM platforms epitomize the Service Network approach due to their transparent support for multiple different protocols with minimal protocol conversion or adaption. Additionally, the MSPP and R-CWDM platforms enable non-intrusive performance monitoring of various protocols, thus supporting SLA requirements. The MSPP and R-CWDM platforms are also complementary they can be used to address different parts of the carrier and enterprise networks or can be integrated to add more capacity and/or finer granularity where needed. In its simplest form, the R-CWDM platform is little more than a smart media converter each node adding only nanoseconds of delay to a Fibre Channel or ESCON signal for example. The MSPP and CWDM and OEO Transport Architectures - IEE Paper.doc Page 11

12 the R-CWDM platforms with GFP-T enhancements (T = Transparent option), add extra cost and transmission delay due to the extra processing, STS/VC switching and OC-n / STM-n multiplexing functions involved. However, this extra GFP processing, switching and multiplexing overhead is relatively small, less expensive and adds significantly less delay compared to alternative ATM, Gigabit Ethernet, IP and MPLS protocol conversion options. The result is that for all intents and purposes, the intrinsic xwdm and GFP transport options more closely replicate native protocol transport than any other transport standard and thus are most suited to supporting Service Networks. 5. Future xwdm and 10G OEO Transport Options This section focuses on recent technology trends and their impact on the future of xwdm and OEO transport platforms Hybrid CWDM and DWDM Multiplexing As shown in Figure 1, two of the CWDM bands (centred on 1531nm and 1551nm) overlap with the most commonly used DWDM band (the C-Band), which is generally defined by the gain profile of most EDFA amplifiers. The successful but sometimes reluctant adoption by carriers of CWDM technology has left them with a fear that their installed base of G.652 fibre could only practically support 8-12 CWDM wavelengths which might be exhausted too quickly. Additionally, there has been a question mark over the ability of CWDM to support 10G protocols in the future for at least IR distances. To-date, 10G CWDM has been focused on VSR distances [4]. To combat these fears, several CWDM equipment vendors have publicized their support for hybrid CWDM and DWDM multiplexing. This would involve for example, using all CWDM wavelengths other than 1531nm and 1551nm initially, and then when more xwdm channels are required, using the DWDM wavelengths in the 1531nm and 1551nm bands. This scheme is illustrated in more detail in Figure 10. As can be seen, up to 32 additional wavelengths with 100GHz spacing can be supported in this way. This enables a low cost CWDM entry strategy and keeps the door open for the higher-cost DWDM interfaces if and when needed. The recent announcement of DWDM SFPs with 100GHz wavelength spacing and the recent formation of a DWDM SFP MSA forum [1] add support for the hybrid CWDM and DWDM multiplexing scheme. However, such SFPs will initially only be available at rates up to 2.7 Gbit/s. As discussed previously in section 1.2, 10 Gbit/s will most likely be available in the XFP package and it is more likely that XFP will focus on fixed and tunable DWDM coloured optics, rather than CWDM. This is consistent with the hybrid CWDM and DWDM multiplexing scheme. Figure 10 DWDM Wavelengths in the 1531nm and 1551nm CWDM Bands CWDM and OEO Transport Architectures - IEE Paper.doc Page 12

13 One of the challenges to hybrid CWDM and DWDM multiplexing is the additional filter loss that is introduced and the impact that this will have on the size of all-optical OADM networks. As outlined in section 3.3, all-optical CWDM OADM networks already have significant size limitations due to accumulated loss and dispersion. Adding additional hybrid CWDM/DWDM filters will only accentuate this problem. Applying EDFA gain in the DWDM band is not really a viable solution in this case, since the CWDM wavelengths also need to be amplified. As mentioned in section 3.3, inserting 3R regenerator cards where and when needed is one solution that works for both CWDM and DWDM. However, in the case of CWDM networks, this was demonstrated to be a sub-optimal solution compared to the R-CWDM approach. The question is, can the R-CWDM approach with its integrated array optics implementation be cost-effectively extended for DWDM wavelengths (R-DWDM) and 10G protocols? The answer to this lies in the recent announcement of new generation array optics technologies that are highly integrated using Indium Phosphide (InP) substrates. These are referred to as Photonic Integrated Circuits (PICs) [5]. PICs integrate multiple DWDM lasers, modulators, wavelength multiplexers, demultiplexers, and photo-detectors. These devices, each smaller than a pencil eraser are claimed to be capable of transmitting and receiving 10 DWDM wavelengths, each at 10 Gbit/s, over long spans of optical fiber. As evident from Figure 10, 20 DWDM wavelengths are readily available in the 1531nm and 1551nm CWDM bands to support a SFW implementation of a 100 Gbit/s DWDM OEO platform. As shown in Figure 11, a new R-DWDM platform can be readily added to an existing R-CWDM node using a hybrid CWDM/DWDM filter module. This can happen without causing any more than 10ms disruption to existing protected services. All the features and benefits of 10 Gbit/s R-CWDM platforms can be adopted by the new generation 100 Gbit/s R-DWDM platforms. 1 Fibre 1 Fibre West Filter Passive Optics East Filter 8λ CWDM Any 8 of Wavelengths 8λ CWDM x DWDM Wavelengths in nm CWDM Bands R-CWDM 50M 2.5G Client Services OEO using Regenerative CWDM Architecture x DWDM Wavelengths in nm CWDM Bands Next Generation R-DWDM OEO using Regenerative DWDM Architecture 8-16 x 10G DWDM Channels with Array Optics 2.5G 10G Client Services 8-16 x 10G DWDM Channels with Array Optics Figure 11 R-DWDM upgrade to an existing R-CWDM node Gbit/s ADM Technology In support of next generation 100 Gbit/s MSPP OEO platforms, VLSI integration has evolved to the point that silicon substrates can accommodate a 10 Gbit/s ADM on a chip [6]. Further integration is expected that will include some or all of the following: multi-protocol GFP mapping; Ethernet MAC and Resilient Packet Ring (RPR) MAC. The integration of new generation, 100 Gbit/s DWDM array optics and MSPP OEO functions into a single platform would be a natural evolution of these technologies. CWDM and OEO Transport Architectures - IEE Paper.doc Page 13

14 6. Conclusion This paper has outlined the evolution of carrier optical networks from their 1 st generation OEO implementations through 2 nd generation all-optical implementations to what could be described as 3 rd generation OEO implementations. In parallel with this evolution, has been the evolution towards Service Networks which transport and manage multiple protocols with minimal conversion relative to their native format. It is now recognized that different protocols are optimized for their respective service requirements and should not be unnecessarily tampered with. Previous generations of broadband networks, such as ATM, required excessive protocol conversion (adaption) adding significant delay and cost to the transport of multiple protocols. The expected evolution away from all-optical networking to 3 rd generation OEO networking is due to both market and technology forces. In terms of market forces, the carriers have recognized that their highest margins are derived from value-added services that rely on systems with electronic multiplexing, switching, grooming, processing, monitoring and management capabilities. The much vaunted bandwidth explosion has not been accompanied by a similar trajectory in carrier revenue. Factors such as the drastic reduction in price per bit and adoption of lower margin data services by end users is forcing operators to rethink their traditional business model. In terms of technology forces, the development of low cost CWDM based array optics, VLSI based GFP processing and 2.5G SFP interface technologies has enabled the development and deployment of various 10 Gbit/s OEO platforms that support multiple services. More recently, the announcement of low cost DWDM based array optics (Photonic Integrated Circuits), 10G ADM on a chip technologies and 10G XFP interface technologies are expected to enable the development and deployment of various 100 Gbit/s OEO platforms that support multiple services. 7. References 1. DWDM FSP MSA Forum, 2. R. Halgren, RBN, Characteristics of CWDM: Roots, Current Status and Future Opportunities 20 Sep R. Halgren, RBN, CWDM and GFP in the Metro Core 5 Feb Optical Internetworking Forum (OIF), Very Short Reach (VSR) Level 5 Project for 40 Gbit/s optical interfaces, Feb Infinera, PIC technology, 6. Parama Networks, 10G ADM technology, 8. Acknowledgements RBN acknowledges the copyright of this paper belongs to the IEE in Conference Publication FUTURE CHALLENGES AND OPPORTUNITIES FOR DWDM AND CWDM IN THE PHOTONIC NETWORKS IEE Midlands Communications Group, 11 th June 2004 CWDM and OEO Transport Architectures - IEE Paper.doc Page 14