Why consider metropolitan DWDM?

Th 08spr12 1

Dense wavelength-division multiplexing (DWDM) offers carriers the potential for vast increases in the availability of bandwidth. Until recently, the focus of DWDM application has been almost exclusively on the national and international backbone routes, where very large circuit cross sections can be aggregated. However, the potential use of this technology in metropolitan areas has generated significant discussion in the last year. Major telecommunications equipment vendors such as Fujitsu, Nortel Networks, Alcatel, CIENA, and Lucent Technologies have announced metropolitan DWDM products. And incumbent local-exchange carriers (ILECs) in major metropolitan areas, such as Bell Atlantic and BellSouth, are investigating the technology and conducting DWDM trials. But although much has been said about the metropolitan market in general, very little has been said about the justification for metropolitan DWDM.

On the surface, metropolitan DWDM systems appear the same as those serving long-haul backbone routes. Both use the technique of optically modulating a single fiber with many independent light signals, each at a different wavelength. There are differences, however.

The few metropolitan DWDM systems available today

  • support 16 optical wavelengths, although some are capable of 20, 32, or even 40
  • carry at transmission rates of OC-48 (2.5 Gbits/sec); some systems support OC-192 (10 Gbits/sec).
  • offer ring architectures with standard protection schemes (there is a lot of variability on this issue-some vendors do not support ring switching)
  • do not interoperate with other vendors' equipment
  • are specifically intended for the short-haul, local environment.

This last characteristic is a fundamental difference between long-haul and metropolitan DWDM systems. Long-haul DWDM systems are designed for applications in networks that span hundreds, even thousands, of miles, while metropolitan DWDM is intended for much shorter distances. For example, BellSouth recently adopted metropolitan DWDM technology from CIENA to interconnect wire centers located 15 miles apart. The ILEC is also selecting a metropolitan DWDM vendor for offices less than 5 miles apart.

These shorter-distance parameters allow DWDM vendors to build systems based on less-stringent specifications and reduce system costs. The fact that today only proprietary technology is available for metropolitan DWDM also lowers system costs.

As these factors imply, cost is a fundamental issue with metropolitan DWDM. In the metropolitan environment, a lot of the traffic is viewed as not providing revenue. Cost is therefore much more critical than in the long-haul market, where virtually all the traffic generates revenue.

Now look at the application for which these new systems have been designed. The data and averages used in this application model are based on published information from Bell Atlantic and Nynex, except where noted.

Our metropolitan area network model will be designed for a city of one million people. Using an average of three people per household, we can estimate this city contains 333,333 households. Given a penetration rate of 95% and an average of 1.2 phone lines per household, the model network must accommodate 380,000 residential lines. Using a ratio of 3:1 for residential to business lines, this area also would have 187,164 business lines, or a total of 567,164 lines.

Such a city is typically served by a number of central offices (wire centers), each with trunking to all of the others to allow local calling. (In practice, metropolitan areas of the size discussed here use one or more interlocal tandems with direct "high-usage" trunks between each pair of central offices. This refinement is ignored in this analysis because it has only a minor impact.)

From the author's experience (and general industry norms), we will apply the following parameters to the model:

  • a fill ratio of 85% (the ratio of working to installed lines)
  • an average size of 30,000 lines for each wire center
  • an average busy hour traffic per line (two-way average of business and residence) of 6 ccs (1 ccs = 100 call seconds)
  • a ratio of 50% local (within the wire center) traffic
  • a ratio of 40% of total trunks for specials (various types of private lines) going to other wire centers.

These parameters will result in a model for the metro area with the following characteristics:

  • 23 wire centers-each of three dial units
  • total traffic per wire center of 153,000 bhccs (busy-hour ccs)
  • total traffic from each wire center to all of the other wire centers of 38,250 bhccs
  • 22 trunk routes from each wire center going to all the others
  • each trunk route will have 165 trunks and "specials" (which include 59 switched trunks in each direction plus 47 private lines)
  • 53,737 total trunks and specials (all routes)
  • a total "cross section" of trunks and specials to or from each wire center of 3634.

Note that 23 times the cross-section trunks does not equal the total trunks (53,737) because the nonprivate-line trunks have two appearances-one at each of two wire centers. The "total trunks" is equal to the private lines plus the traffic trunks. It is assumed that all private lines go "off-net."Th 08spr12 1

Fig. 1. This metropolitan area network model illustrates a portion of a large mesh network with each node having an array of trunk routes to each of the other nodes. The entire network would seem to require a significant amount of capacity-but is it enough to require metropolitan DWDM?

The resulting model is of a large mesh network with each node having an array of trunk routes to each of the others. Figure 1 illustrates a portion of this network model to clarify the view.

Comparing the typical metropolitan system described above with the parameters contained within our metro model doesn't present a glowing case for the technology. For example, a metropolitan DWDM system with 16 wavelengths (in each direction) operating at OC-48 can support up to 32,256 voice trunks per wavelength (DS-0s transmitting at 64 kbits/sec) or a total of 516,096 voice channels. These capabilities seem like overkill when our model requires only 3634 trunks at a given node (less than two OC-3s, each carrying at rates of 155 Mbits/sec) and individual routes of only 165 trunks (less than a DS-3 or 45 Mbits/sec).

Given this comparison, what is the business case for metropolitan DWDM? "None," some analysts have said-and they may be right. But rather than endorse this conclusion, here are some situations that may justify metropolitan DWDM. These reasons fall under two categories: "plant extensions" and "competitive reasons." Plant extensions represent the more traditional rationales for DWDM. Network growth and cost factors represent the principal drivers for including DWDM in the "plant extension process" of incumbent carriers. Here are some extension scenarios.

Fiber/duct exhaust. Still the most obvious application and the one most likely to drive this market in the short term, the use of DWDM on existing fiber not only results in duct relief, but also allows system administrators to reclaim fibers for other purposes. If metropolitan DWDM is deployed in an existing route, it is possible to transfer trunks from multiple fibers to the new system and thus free fibers for use with other services or for lease to other providers.

Obsolete electronics. Many fiber-transport solutions used for local interoffice facilities in the metropolitan plant are now quite old. Some do not interface with modern operations, administration, maintenance, and provisioning (OAM&P) systems and are not compatible with other transport solutions. Replacement of obsolete equipment of this type easily justifies the consideration of metropolitan DWDM.

New wire center development. The need to develop a new wire center requires the introduction of trunk routes, which presents an excellent opportunity to review the entire metropolitan inter office network for the use of DWDM.

OC-48 plus DWDM versus OC-192. If there is simply a capacity problem, one alternative is to increase the Synchronous Optical Network/Synchronous Digital Hierarchy (SONET/ SDH) level from OC-48 to OC-192. Some studies indicate an economic advantage to OC-192 over multiple OC-48 wavelengths (see Lightwave, December 1998, page 92). It is not clear from this research whether there is a lower initial cost for an OC-48 metropolitan DWDM solution as some vendors claim. Still, this option deserves careful consideration.

Beyond forecast traffic. If an ILEC has entered the broadband-to-the-home race, the carrier may need to contend with "beyond-forecast" traffic. Let's say the ILEC is offering digital subscriber line (DSL) circuits. Looking just at traffic on digital-loop carriers (DLCs) can provide an idea of the traffic level. Currently, estimates for the amount of residential traffic served from DLCs range from 30% to 50%, and it takes typically three or four DS-1s to handle the traffic on a 500-line DLC. If 30% of the customers on a typical DLC were to take asymmetric digital subscriber line (ADSL), then the required trunking back to the wire center would be a minimum of one OC-3-and more realistically, two OC-3s-based on a mere 1-Mbit/sec bandwidth for each of the subscribing ADSL customers.

Providing one OC-3 would represent a 25x increase in bandwidth over the four DS-1s now required; two OC-3s would mean a 50x increase. Most, if not all, of this added traffic would be dumped directly onto the interlocal networks serving an Internet service provider (ISP) office. If 30% of the residential lines are served (with 30% subscribing to ADSL), then the metropolitan-area model's 3634 trunks per wire center cross section jumps to 16,422 effective trunks-a 350% increase. Such an increase would require drastic increases in the interoffice facility routes-a situation that clearly suggests the need for a careful investigation of the metropolitan DWDM option.

In addition to these physical extension issues, competitive requirements also may make metropolitan DWDM attractive. While the distinction between physical extension and competitive factors may not be pure in every instance, the following drivers tend to be more market driven.

A desire to "stockpile" bandwidth ahead of demand. Some emerging long-haul carriers have followed this strategy, and they have used DWDM to do it. Once established in the market, the carrier can also provide bandwidth to others entering the market-even competitors-because the cost structure will allow the provider to offer capacity more cheaply than new entrants can build it.

Advanced transport services. One of the advantages of using metropolitan DWDM is that the individual wavelengths are independent of each other. A single wavelength can carry traditional SONET/SDH, while others carry different transport protocols, which offers great flexibility in terms of service capabilities, particularly with the higher-bit-rate applications. Services such as direct-to-fiber Internet protocol (IP) with terabit routers, Enterprise System Connection (Escon), D1 video, or gigabit IP all require greater than SONET/SDH rates but can be carried at native rates on metropolitan DWDM.

Reliability. Although not currently available in all vendors' offerings, some metropolitan DWDM products provide ring-protection switching capabilities. Interlocal networks that cannot guarantee continuation of service face a severe competitive disadvantage in the market.

To improve flexibility and operational characteristics. The interlocal network in a metropolitan area is a very large, complex mesh. Such a network, with its many different trunk routes (note the 22 different routes from each of 23 wire centers in the Figure 1 model), is very difficult to maintain and administer. Changes to the metropolitan area network are frequent and often require action at many different points. Because end customers move, grow, and alter their service requirements, rapid change is the norm for these networks. Th 08spr12 2

Fig. 2. In this metropolitan DWDM architecture, one wavelength is routed from each of the wire centers to the digital crossconnect system (DCS). All the crossconnection is handled through the DCS. Many long-haul carriers use this type of architecture.

The metropolitan DWDM architecture mitigates this problem and provides excellent responsiveness to changing customer requirements. Metropolitan DWDM systems enable the assignment of a wavelength to each wire center. The termination of all the wavelengths is at one wire center on the digital crossconnect system (DCS). Figure 2 shows a metropolitan DWDM architecture supporting a high-capacity DCS. With this architecture, the routes are reduced (effectively) to one from each of the wire centers to the DCS. All crossconnection is handled through the DCS. Virtually all changes can be made from the DCS using computer commands. Large long-haul carriers use this type of architecture.

To facilitate local enterprise networks and connection to long-haul enterprise networks. A key concern for enterprise networks is survivability. Some network planners have gone to great lengths to ensure survivability in almost any circumstance using separate entrances, different carriers, mirrored sites, and other precautions. The ability to offer a true ring-protected service is key to competing in this service area. Metropolitan DWDM can immediately meet this requirement.

Most enterprise networks are more than local in nature. A concern of the long-haul carriers is how these enterprise infrastructures will be terminated in the local end points. Metropolitan DWDM rings offer the local carrier an opportunity to partner with other carriers in the provisioning of these wide-area enterprise networks.

While many reasons to deploy metropolitan DWDM are compelling, it is still an emerging technology, and DWDM technology purchased today could be obsolete by the time it is placed in service. Waiting for the ultimate DWDM product is not the solution, however. Careful evaluation of DWDM system technologies that meet a network's requirements must also involve consideration of the product's life cycle and maturity. u

Clifford Holliday is the president of B & C Consulting Services (Colleyville, TX).

More in Network Design