Cost reduction and return on investment are clearly the watchwords in the current depressed optical networking market. In this market, it is critical to focus on methods of achieving cost reduction in optical networks, and to quantify the savings that can be achieved through each method.
The costs incurred by optical-network system vendors and service providers can be lumped into two main categories, capital expenditures (capex) and operating expenditures (opex), with significant potential cost savings in each category.
Many of the cost-reduction opportunities for network build-outs and upgrades relate to terminal equipment cost reduction. Currently, terminal equipment costs may constitute as much as 60% of total network costs, and this percentage will only rise as wavelength division-multiplexing (WDM) penetrates formerly single-wavelength links, and as channel counts increase.
The network example used throughout this article is a simplified 25-node regional WDM network, assuming only 10-Gbit/s ports on both client and line side. The network contains 16 WDM channels. However, the basic principles and trends in cost reduction used in the example apply to any network configuration.
The concept of a savings multiplier is a useful concept for analyzing cost-reduction effects. This measure simply translates a reduction in equipment cost, size, or power into system cost savings.
In our simple example above, reducing the cost of a line-side line card (assuming both transmitter and receiver are on one card) by $2000 implies a network cost savings of $800,000 (a savings multiplier of 400), while reducing the cost of a client-side line card by $1000 could result in an additional $400,000 of savings, for a total of $1.2 million. As the number of nodes, wavelengths, and networks increases, the total cost savings on terminal equipment also increases.
In the past few years, there has been a significant trend toward outsourcing optical interface hardware development to component companies through the use of standards-based transponder modules in optical terminal equipment. Transponders offer cost reduction from a combination of component/subsystem cost reductions and the reduced development costs associated with outsourcing. Table 1 shows two examples of line-card architectures using components that are readily available in the marketplace, with a rough cost comparison between discrete and transponder-based line-card design options.Note that these are, to a large extent, "best achievable" numbers for a discrete design. Current-generation discrete 10-Gbit/s WDM line-card costs can run as high as $75,000, of which approximately $45,000 represents the cost of the optical interface hardware itself.
The comparison between an internal, discrete development and an outsourced, transponder-based model is fairly even (depending on yield assumptions) if we look only at the yielded bill-of-material (BOM) cost of the line card as a comparison point. However, the real benefits of the outsourcing model become apparent when product development and manufacturing costs are factored in.
The fixed or overhead costs, as shown in Table 2, are related primarily to labor and capital expenses. This analysis is likely a conservative estimate as it does not include items such as initial hiring costs, administrative support costs, and so on.Only vertically integrated transponder suppliers with internal IC and optical components capability and a low-cost manufacturing infrastructure will be able to consistently achieve the low internal costs necessary to make this business model work. Note that transponder suppliers with a broad product line serving many customers are more likely to achieve the higher volumes needed to amortize development costs than internal development efforts that serve only one company.
In addition to reducing overall cost, outsourcing essentially converts development and manufacturing expenses from a fixed cost into a variable cost; equipment vendors pay development costs only on what they sell. Flexibility is also increased through the ability to source standardized transponders from multiple vendors.
Transponders can reduce cost in operational expenditures by savings in space, power, and sparing:
Space Savings. Ever-tightening space constraints have made shrinking the footprint of optical networking equipment a crucial goal. Some of this can be accomplished at the architecture level by simply collapsing optical equipment into multipurpose boxes such as MSPPs.
At the board and component level, space savings can be achieved by: moving from RF boxes to chips mounted directly on the circuit board; greater electrical, optical, and optoelectronic component integration; and reducing the fiber routing footprint through optical integration, bulkhead optical connectors, and blind-mate optical line-card connections to eliminate fiber pigtails.
Space savings from component technologies are more difficult to analyze, as the space in terminal equipment is quantified in units of line cards, shelves, and racks. However, if we take a realistic case in which the maximum number of cards per rack before space reduction is 12 (4 per shelf and 3 shelves per rack, again assuming 10-Gbit/s cards), and the redesign of a line card enables a transmitter and receiver to be integrated together on one line card instead of separate cards, resulting in savings in our regional network (see Table 4, top).
Power Savings. The main benefit from reduced DC power consumption is the space savings discussed above because increased density without reduced power dissipation results in unmanageable thermal requirements for shelf designs. Increased levels of IC and optoelectronic module integration, as well as new technologies such as uncooled lasers, enable this decreased power dissipation.
Power reduction also offers cost savings in its own right; however, this savings turns out to be fairly small. A power reduction of 2 W on the line side and 1 W on the client side results in a total network cost savings of just $1050/year from power alone (see Table 4, center).Sparing Savings. For systems with multiple wavelengths and/or multiple services/protocols, reducing the number of spares required at each site can offer significant savings. Assuming an $8000 WDM transponder price, then the network savings from sparing for a transponder offering full C-band wavelength tuning and multiple protocol support could be as high as $6.2 million on the line side if the number of spares required is reduced from 800 to 25 (see Table 4, bottom). Inventory management costs will also be lower in this case.
Several trends offer clear opportunities for significant cost savings, including integration, expansion of the installed fiber capacity, and automated manufacturing. Further component and subsystem integration will decrease costs through increased line-card densities. Opportunities include:
- Chip integration. More highly integrated ICs will reduce chip counts by, for example, integrating mux/demux, forward-error correction, and framing functionality into a single chip.
- Optoelectronic integration. Modules will continue to shrink to smaller form factors, even as more functionality (such as laser + driver) is integrated into a single package.
- Photonic integration. As planar lightwave-circuit technology matures, multiple optical functions (such as mux/demux, monitoring, attenuation, or switching) will be integrated into a single optical module.
Better use of the installed fiber base will remain a priority over the next several years. Wavelength-division multiplexing will continue to penetrate metropolitan area networks, increasing the carrying capacity of existing fiber. Newly maturing technologies such as electrical dispersion compensation and duobinary modulation will help to raise the transmission rate and reach limits of fibers currently limited by polarization-mode and chromatic dispersion.
Finally, though true mass production will have to await a market recovery, transitioning component manufacture to mass-production techniques is more critical than ever to reduce component costs through lower labor costs and increased yields.
Scott Schube is senior technical marketing engineer for DWDM products in the Optical Platform Division of Intel, 8674 Thornton Ave., Newark, CA 94560. He can be reached at [email protected].