Please see CIR's report Wave Division Multiplexing, Photonic Switching and the Coming of All Optical Networks 1999

Since mid-1998, the DWDM industry has begun to focus on the opportunities in the "metro" segment of the network. This is the segment between the access network and major backbone facilities. CIR believes there are several reasons why the metro segment is getting so much attention.

• The long-distance DWDM market has shown some signs of early maturity. Deployment of DWDM by the major long-distance carriers is not accelerating and, more importantly, the largest long-distance networks are beginning to settle on a few DWDM suppliers, creating barriers to entry for others, especially start ups. For this reason, DWDM vendors have started to look around for new opportunities.

• The metro segment appears to offer such opportunities. It is potentially a huge market, because it already consumes billions of dollars of TDM (SONET and T-carrier) equipment every year. DWDM vendors argue that if they could just grab a little of that market, they could do very well. And the metro segment does appear to be in need of such bandwidth, since it is squeezed between the access network÷whose bandwidth is being boosted by xDSL, cable modems and ATM access devices÷and the long-haul backbones that are being pumped up with DWDM.

• As discussed below, an argument can be made that the technical requirements for a metro DWDM box are significantly different from those of a long-haul DWDM box. This means that established DWDM suppliers who have previously made their living by selling into the long-distance network may not have significant competitive advantages over new entrants. This is the economics on which the emergence of companies such as Lightera (now part of Ciena) and Sycamore and Tellium are predicted.

CIR believes that the analysis given above is substantially correct. The metro area will need more bandwidth. But when? And will DWDM be the way to provide that bandwidth? For a few service providers÷perhaps as few as 20÷the answer to these questions appears to be "immediately" and "yes," respectively and they are already deploying DWDM in the metro area. These service providers are mostly CLECs deploying in green field sites. But the ILECs, in whose networks the greatest potential for DWDM sales clearly lies, are mostly dabbling with DWDM.

CIR believes that, today, many service providers are recognizing that DWDM is the leading edge technology, but are finding that cost and the immaturity of this technology for metro applications are still driving them towards a fiber/SONET solution. The speed and degree to which these interrelated problems can be overcome will determine the size of the metro DWDM market over the next few years.

CIR believes that there are serious cost-related factors that are delaying÷and will continue to delay÷the mass introduction of DWDM into the metro environment. These are the following:

• Inappropriate engineering of DWDM systems. Until very recently, LECs have found that the DWDM systems available to them have been primarily designed to meet the needs of long-distance providers. These were usually too expensive because they supported data rates higher than required by metro area service providers over distances longer than required by these service providers. This problem is going away quickly, because more and more vendors are bringing metro DWDM systems to market. These systems are less costly because they do not need the amplifiers used for long-haul applications and because they can use lower cost transceivers for lower data rates. On the other hand the special features that are required of DWDM systems÷see below÷add to the costs.

• The "low" cost of laying fiber. Metro DWDM is being touted as a solution to the high cost of pulling in new fiber in the metro environment. However, pulling in new fiber may not be all that expensive. It certainly is expensive in the largest metropolitan areas, where DWDM may be a way around crippling labor costs and regulations associated with laying new fiber. For example, in New York work of this kind must be done at night with union labor, which is enormously expensive ö as much as $20,000 per km. Assuming that the span between nodes is about 20 km, service operators would be faced with spending $400,000 on pulling in new fiber plus the cost of a SONET ADM. Or they could buy a DWDM system. In such cases the choice is clear÷the DWDM system is almost always the winner. But New York is highly untypical. In many instances, $1,000 per km to pull fiber if existing conduit is in place may be a more accurate figure. Even with the additional cost of a SONET ADM factored in, DWDM at current prices would not make economic sense. With the cost of labor relatively static, CIR believes that the metro DWDM systems may have to come down in price by 40 to 50 percent, before we see mass deployment. This will not happen for a few years.

• No regenerators required. In long-haul systems the economic justification for DWDM is frequently the elimination of regenerators. However, most metro systems do not need such regenerators in the first place. Metro area links top out at 120 km÷Nortel research indicates that 90 percent of rings in the U.S. are under 120 km in circumference÷with long-haul networks running up to 1,000 km or more. Regenerator elimination is a key reason why DWDM is deployed in the long haul÷it never will be a reason why it is used in the metro area.

CIR believes that even an overnight drop in the price of metro DWDM equipment would not lead to a scramble by service providers to deploy DWDM. Other technical issues will have to be addressed too. These relate to the kinds of services that they must support and also to the important issue of general network management. As far as the latter is concerned, there is a general consensus that metro DWDM lacks the path restoration, management, and time-slot assignment capabilities available in SONET ADMs. Issues that metro DWDM vendors are currently contending with are:

• Need to support standard LAN internetworking services/interfaces. At the backbone level, differing services have mostly been encapsulated into SONET frames, so that the primary concern is with equipment that has open interfaces to established SONET products. However, at the metro level, networks often support services in native form. In particular, CIR believes that service providers will require that metro DWDM platforms be able to support native LAN protocols to enable them better to generate revenues from high-end LAN internetworking services. CIR believes that metro DWDM systems should be equipped with a port for Gigabit Ethernet and possibly Fast Ethernet. This is not hard to do and some systems are doing this now. A few also support FDDI.

• Need to support channel services/interfaces. In addition to "mass market" LAN interfaces, metro DWDM systems may require to support the high-end channel interfaces÷Escon, Ficon and Fibre Channel÷associated with the large data centers in financial institutions, utilities and government. It is such centers that are likely to be the main target for services provisioned over metro DWDM platforms and DWDM systems seem much better suited to supporting deployment of services operating under these protocols that run at 200 Mbps to 1.06 Gbps than the conventional SONET infrastructure with its limited bandwidth and fixed data rates. Relatively few metro DWDM systems currently support such protocols. And, as with LAN services/interfaces, a lot more is involved here than just supplying physical interfaces. Vendors must provide billing solutions for those service providers who want to provide high-speed services of this kind.

• Path restoration. SONET rings are either UPSRs (unidirectional path-switched ring), or BLSRs (bidirectional line-switched ring), as defined by the Bellcore (Telcordia) standards. UPSRs offer 1:1 redundancy throughout the ring; BLSRs provide a degree of redundancy through some extra capacity inherent in the SONET standards. Currently, few metro DWDM vendors have been able to provide equivalent functionality and CIR believes that the ability to do so will put vendors at a major competitive advantage. One metro company that has specifically recognized this is Sycamore, and the industry consensus is that path restoration for DWDM equivalent to that of SONET should be widely available within a year to 18 months.

• Protection Switching. This is the issue that is constantly raised as a key advantage that SONET has over DWDM. However, CIR believes that this problem may be somewhat exaggerated. Based on interviews conducted for this study, CIR believes that the disparity between SONET and DWDM will be successfully addressed within about a year. In addition, the problem appears to have less to do with DWDM than the fact that for the foreseeable future, DWDM will be used in conjunction with SONET. In an entirely DWDM environment, most DWDM devices would shift over to the backup fiber in about 25 milliseconds or less, whereas SONET ADMs can take up 50 milliseconds. But where SONET and DWDM are used in conjunction, the DWDMâs intrinsic protection switching are turned off to avoid conflict with SONETâs system. This in inefficient and results in a slower than SONET response.

• Thus the protection switching argument that is used to project that DWDM will continue to have deployment problems could just as easily be used to be argue that there is a need to move quickly to DWDM-only metro systems. However, with the huge embedded base of SONET and the commitment that the Bell companies have made to SONET, CIR does not believe that pushing DWDM-only solutions in the metro segment would be a successful competitive strategy at this point in time, although it may be within a few years. However, significant competitive advantage is likely to be enjoyed by vendors who can capitalize on DWDMâs intrinsic protection switching capabilities. Sycamore, which claims a fail-over rate of 3 milliseconds, is one company that has recognized this.

• Time slot assignment. Yet another advantage claimed by SONET is time slot assignment. SONET ADMs are programmed to assign a time-slot from one customer to another. Nortelâs metro DWDM solution, using the technology it has acquired from Cambrian, contains an electrical crossconnect that reassigns wavelengths between customers. The only alternative appears to be to physically moving cards in the DWDM system chassis, which harkens back to the days of the old patch panel and is hardly an acceptable solution in the AON era.

CIR believes that the usual experience curve, Mooreâs Law and mass production effects will bring down the cost of metro DWDM over the next two to five years to a point where metro DWDM begins to make economic sense for many service providers. In addition, we would also expect most of the technical issues to be resolved in that time frame.

CIR believes these changes alone will result in a significant penetration of the metro area by DWDM. But we also believe that there is no short-to-medium term threat to SONET in this segment of the network÷SONET will continue to grow at the expense of T-carrier. However, CIR believes that SONET may ultimately begin to be threatened as bandwidth demands from larger business users and other service providers put competitive pressures on service providers to rapidly provision very high bandwidth pipes. CIR believes this functionality will be a key to success in the high end of the market and may serve as a distinguishing feature for CLECs and ISPs aiming at the more sophisticated user.

CIR believes that, while the need for rapid provisioning of high data rate pipes in the metro area may take two or more years, before it has a significant impact, it could be crucial in DWDM scoring a victory over SONET in the metro area, although with the huge installed base of SONET, which gets bigger by the day, that victory will be slow in coming. DWDM architectures will score points over SONET in the coming provisioning wars.

In the classic SONET architecture, central offices serve larger corporate and service provider customers over access rings that are built out from each central office. These typically operate at OC-3 or OC-12 and survivability is guaranteed by building the rings with two fibers. The rings serve about four nodes and these nodes are served by SONET ADMs located at the customer premises. Generally, these nodes provide the customer with DS1, DS3 and OC-3 services, the DS1 and DS3 traffic being dropped into SONET frames for transport on the SONET net.

Unfortunately, this architecture will not easily support the future demands for OC-12s by larger customers, since todayâs SONET rings are already quite busy. This means that service providers must either build a new customer ring from scratch, which usually takes more than six months. Or it can merely upgrade its SONET equipment, which still takes several months. DWDM apparently offers an alternative to such delays. DWDM rings from the central office offer much higher data rates to the customer premises and DWDMs can also be installed at the customer premises equipped with cards that allow customers to add or drop wavelengths that can carry much higher data rates than conventional ADMs. Thus metro DWDM will enable service providers to be much more responsive to the needs of their largest customers.

CIR believes that the introduction of metro DWDM is most likely to occur in evolutionary fashion and that the first stage of this evolution would still retain SONET as a major part of the solution. In this first stage, DWDMs would serve customer premises SONET ADMs. As bandwidth demands by customers begin to increase, the SONET ADMs would be replaced by WDMs with services provisioned directly over the optical layer.

This evolutionary progression from SONET to DWDM in the metro environment will, CIR believes, help boost the DWDM market significantly, but a revolutionary change will occur as service providers learn to use DWDM in the metro area in conjunction with optical crossconnects to offer what is currently referred to in the industry as "just-in-time provisioning of bandwidth," and which has been known in the past as bandwidth-on-demand. In theory, at least, optical cross connects that route wavelengths from one fiber to another; when combined with metro DWDMs, could enable service providers to assign an OC-48 to a customer in seconds. If a significant market for such services emerge, CIR believes they could ultimately prove the death for SONET serving the large business customer. However, 1) it will be several years before OC-48s begin to sell en masse and 2) there will probably still be lots of other work for SONET to do if it is forced out of this part of the network.

While the cost reduction impact of DWDM is paramount at the present time, CIR believes that the economics of metro DWDM is about to be changed as service providers begin to regard it more as a potential revenue generator. This could occur in two ways.

• DWDM as a multiservice platform. DWDM enables service providers to offer corporate customers a mix of services over a single fiber, but with each service running over a different wavelength. For example, the service provider may offer frame relay on one channel, Fast Ethernet on an another and Escon on yet another. While there is an element of cost reduction in this approach÷the service provider could obviously offer the same services by running multiple fibers into the customer premises÷DWDM essentially provides service providers with a way of adding value to dark fiber services.

• Just-in Time Provisioning. This is very rapid provisioning of the kind that was once referred to as bandwidth on demand. It is dependent on the deployment of optical cross connects, but would serve the needs of users such as ISPs and large financial institutions who want to buy bandwidth at off-peak rates. It could also serve the needs of a more general class of users, who may want (say) OC-48 to their Web server for some special event, such as a major sale. In quieter times, such users may need no more than an OC-3.

As DWDM appears to offer new value added revenue generation potential for the service providers, its penetration can be expected to accelerate. DWDM vendors will begin to make a value added pitch of their products to service providers as more than just an alternative to putting in more fiber and SONET/ADMs. We are still a long way from the all optical networks (AONs), but the new metro DWDM platforms are a significant step towards the optical future.