Is too much fiber not enough?
SPECIAL REPORTS: Optical Communications Market
Wall Street says there's a capacity glut. Manufacturers and the TIA say there will soon be a shortage. Does it matter who's right when money may mean more than facts?
Anyone with a subscription to the Wall Street Journal knows the financial world's take on optical communications. There is significant excess capacity in carrier networks-and it will be a long time before this excess burns away. As a result, optical stocks will continue to plummet, companies will continue to fold, and the manufacturers of optical fiber will have to wait years for a significant rebound in demand.
An increasing number of voices in the optical community-recently joined by the Telecommunications Industry Association (TIA)-are calling Wall Street to task for, at best, erroneous assumptions or, at worst, plain ignorance and irrationality. Armed with data on historical bandwidth drivers, past fiber and equipment usage trends, and a knowledge of how networks "really work," they claim that far from having too much capacity, carriers in the United States are rapidly running out of bandwidth and may face capacity shortages as soon as this year.
Who is right? It may not matter if carriers have to manage their spending to meet the expectations of financial analysts rather than their network requirements.
Perception and reality
One company clearly exasperated by Wall Street is OFS Fitel (Norcross, GA), the former Lucent fiber and cable division now owned by Furukawa Electric and CommScope. The company assembled a group of journalists and analysts (including this author) at the recent Optical Fiber Communications (OFC) Conference to dispel five "misconceptions" promulgated by financial analysts:
- Bandwidth demand growth is slowing down, perhaps to less than 50% annually.
- Only 3% of active capacity is being used.
- Service providers can just use up the spare channel capacity in their networks until things get better.
- Only 5% of fiber is lit.
- The market for fiber won't come back for years and will come back gradually when it does.
The OFS Fitel representatives say these assertions are based on a combination of inordinately pessimistic assumptions, a misreading of research, and a lack of understanding about how carriers use their fiber infrastructure. For example, the popularity of 3% as the amount of active capacity now in use in U.S. telecommunications networks derives from a Merrill Lynch study from March 2001 that put the amount of capacity used in the United States at an average of 3%. (Its latest report pegs the average at 10.7% for 2002.) Averages, of course, imply that sometimes the traffic is greater. Looking at that 3% figure, for those not aware that network planners must design their networks to handle peak loads, it's easy to assume wrongly that 97% of U.S. networks aren't in use at all.
Similarly, this average encompasses every carrier in the United States; newer carriers with recent overbuilds have a lot of spare capacity, while established carriers with older networks may be nearing exhaust on some routes. A blithe application of the 3% figure across the board could greatly underestimate an established carrier's capacity situation. (Misapplication of that 3% figure irked the researchers at Merrill Lynch as much as anyone; for more, see Lightwave, August 2001, page 25.)
OFS Fitel and other critics of the financial press claim that too often apples are mistaken for oranges and the resulting fruit salad gives an inaccurate picture of the market. Some people study average bandwidth utilization, others have tried looking at the amount of dark fiber in the ground, and still others have examined the number of empty card slots in transport equipment. The result is a swirl of numbers that often get confused with one another, so if someone offers a guess that 5% of fiber has been lit in carrier networks, that number is close enough to last year's 3% average capacity number to seem like a reasonable estimate, even though the Merrill Lynch study actually estimated that 35% of carrier fiber had been lit.
Finally, it sometimes seems that Wall Street forgets that if a network overbuild is brand new, it's not supposed to be approaching exhaust. As Richard Mack, vice president and general manager of market researcher KMI Corp. (Providence, RI), puts it, "When you put the fiber in, you naturally put it in for a long lifecycle where the engineering spec may be 25 or 30 years, but you may want it to last 10 years or 15 years before you overlay. So you are going to have, in the first few years after it's put in, relatively small utilization."
Based on their research, the folks at OFS Fitel say the reality that Wall Street's five misconceptions obscure looks like this:
- Bandwidth demand growth remains high, a statement they back up with statistical trends such as Moore's Law, the growth in kilobytes per dollar of magnetic data storage and random access memory, the number of Internet hosts, and various Internet traffic demand estimates. "We say 80% [annual growth], just to make people less nervous," says Max Nelson, senior manager of strategic business planning at OFS Fitel, implying that he believes bandwidth demand is growing even faster.
- Peak capacity demands mean more than average capacity utilization, since network developers must engineer their networks for peaks. OFS Fitel says that with network traffic becoming ever more data-centric and with data peak-to-average ratios being anywhere from 10% to 15%, over 60% of active channel capacity currently must be made available. Network planners generally want to add capacity when peak utilization reaches 70%.
- Carriers do not necessarily engineer their networks with an eye toward filling every channel on every fiber before they install new cable. Technological obsolescence in both fiber and equipment, the economic advantages of new fiber and equipment, changes in customer requirements, and a desire to keep capacity in reserve combine to chew up infrastructure faster than some Wall Street analysts realize.
- The percentage of lit fiber in carrier networks varies by segment. Historically, networks have operated with anywhere from 30% to 60% of their fiber lit.
- OFS Fitel believes that fiber shipments are driven more by "top-down" economics than "bottom-up" technology or network needs, with the exception of unusual market influences such as deregulation or Internet frenzy. Thus, it more or less follows the economy and grows at 10 times the gross domestic product growth rate for an overall average growth rate of 20% annually. The company also states that fiber growth per year can be derived from dividing the growth in peak bandwidth demand by the growth in the average bit rate per fiber-a combination the company says will work out to 25%.
The third point bears further explanation. "Network deployments and upgrade paths are [based on] economics and not technology," asserts Nelson. Thus, capacity-whether in the form of fiber, line-card slots, or some other manifestation-will be either used or abandoned based on what makes the most economic sense to the carrier. It is possible that a carrier may move from 2.5-Gbit/sec DWDM equipment to 10-Gbit/sec systems on a route for economic reasons that have nothing to do with whether 2.5-Gbit/sec channels remain available. Similarly, the carrier may install nonzero dispersion-shifted fiber (NZDSF) along that route to support the 10-Gbit/sec equipment even though there is still standard singlemode fiber in the ground that hasn't been lit yet.
Thus, fiber and system resources may be "stranded" for economic reasons, and thus fall out of the carrier's pool of available capacity. Ed DeLong, vice president of engineering and network planning at Broadwing Inc. (Cincinnati), corroborates this notion. Broadwing recently completed its nationwide optical network, which features a backbone "express" network capable of supporting 80 channels of OC-192 traffic per fiber. The carrier is in no danger of running out of capacity any time soon on this backbone.
"One of the things I've told folks is that there'll probably be a disruptive technology that's introduced before I actually scale all the way to 80 OC-192s that moves me to the next set of fibers," DeLong says. "To the extent that I move to the next set of fiber pairs, I will probably do that because there is a technology that drives me there, rather than the fact that I've capped out my existing capacity on the fiber I just lit."
OFS Fitel's Nelson theorizes that the advent of 40-Gbit/sec transport may prove to be the kind of disruptive technology to which DeLong refers. Fiber that needed compensation to handle 10 Gbits/sec will likely be completely stranded in a 40-Gbit/sec scenario-and some of the current NZDSF may run into problems as well. Mack sees some validity in this theory.
Long, cold winter
As a result of their calculations, the OFS Fitel team believes that rather than a capacity glut, carriers will face a shortage of bandwidth as early as this winter. The conclusion is based on simple math: Multiply Merrill Lynch's 2002 average capacity figure of 10.7% by a data peak-to-average ratio of 10 and you get a requirement for 107% of today's capacity during peak data transmission periods.
The TIA is ready to back this assertion in a white paper due to be released before next month's SuperComm in Atlanta. "We do believe that probably by the end of this year, definitely in 2003, that there is going to be a shortage to the point that it's going to cause a slowing down of the traffic-slowing down in the sense that you're not going to get the transmission speeds," states TIA president Matthew J. Flanigan. "Some people are going to pick up the phone and not get dial tone."
The TIA will support the importance of technical obsolescence in how rapidly capacity can dry up. "A lot of the fiber that's out there doesn't efficiently handle the speeds we need today," says Flanigan. "A lot of what's out there isn't efficient to use. It could be used, but it's actually more expensive for a service provider to use it, and it would be smarter and less costly in the long run to put new fiber in."
Flanigan also supports aggressive bandwidth growth numbers. "In our market review that we just announced, we continue to show that there is growth on the services side. And it's pretty significant growth-it's typically up in the high single digits, low double-digit numbers. And in the surveys also that we've looked at, the Internet traffic continues to grow between 60% and 80% a year," he concludes.
Not so fast
Such exclamations of impending bandwidth doom certainly contrast sharply with the view from the financial community. However, before concluding that everyone on Wall Street is as unthinkingly negative today as they now appear to have been unthinkingly positive in 2000, several financial analysts are conducting serious research into bandwidth demand and utilization in an effort to provide their clients with informed advice.
Merrill Lynch is one such firm. Having been scorched by the misuse of their March 2001 report, the researchers at Merrill Lynch make it very clear in their April 2002 edition that they are estimating the average utilization of network resources; that such estimations don't account for geographic disparities or compensate for peak traffic loads; that you will never get close to 100% average utilization; and that carriers will engineer their networks to handle peak loads, not average loads. Moreover, the company states that "well over half" of all traffic in 2002 can be classified as data.
Merrill Lynch focuses on average utilization partly because its research is based on FCC data on frequency, distance, and duration of voice calls, and the company feels confident it can estimate the amount of total traffic that enters the network over an extended period of time based on this data. Similarly, the FCC has also tracked infrastructure deployments. By comparing the amount of traffic going through national networks with the amount of capacity on hand, the researchers can estimate the average amount of infrastructure that was used. To forecast how this average may change, they estimate how fast bandwidth demand will grow and compare this figure with how quickly capacity is growing.
Despite the limitations of looking at average utilization in light of their acknowledgement that networks are engineered for peak loads, Merrill Lynch feels such a figure can still be a good barometer of when carriers will need to add capacity. Historically, average network utilization is 15%. If current utilization is less than that, carriers probably won't need to add much bandwidth; if it is more, a bandwidth shortage may be imminent.
The April 2002 report's 10.7% average utilization figure contrasts with 6.6% for 2000. (Last March's 3% figure did not include protection bandwidth; the current figures do.) However, Merrill Lynch pushed out its prediction of when average utilization will reach the magic 15% number until the end of 2003 or the beginning of 2004. The biggest factor in the gloomier forecast is the firm's lowered estimation of bandwidth growth, from an expected rate of 85% in its last report to 50% in the current edition.
The influence of this bandwidth growth figure is enormous. As the Figure shows, if the bandwidth growth rate is 70%, Merrill Lynch estimates that U.S. networks will reach 15% average utilization this year, which would imply that it's not too soon to start adding capacity. The analysts stand behind the 50% figure, the middle of a range of estimates from 30% to 70% annually; Merrill Lynch recently confirmed this figure based on data from the FCC that indicated data traffic grew 48% between June 2000 and June 2001. What's more, they feel that peak-to-average ratios are closer to 5-6% than 10% or 15%.
Not surprisingly, the strategy team at OFS Fitel begs to differ. "If you're telling me that bandwidth demand has slowed, what is causing that? Are people sending fewer e-mails? Are people going to Websites less? Are companies doing less computer-to-computer data transfer, which is really the bulk of stuff?" asks Nelson. "No, they're doing more of that because when they lose people they have to rely more on automated process."
Money is the greater shortage
The fact that bandwidth demand forecasting remains an inexact science was demonstrated by Corning Inc. during its annual press and analyst briefing at OFC. The company showed a chart of approximately 10 forecasts for annual bandwidth growth that ranged from 53% to 150%. Yet, even if the higher forecasts are correct, carriers may not expand their capital-expenditure budgets, because while a bandwidth shortage is a matter of debate, the cash shortage is not. Carrier revenues continue to decline, which makes your average telecom company something of a pariah to investors. In such an environment, borrowing money to add capacity that financiers don't believe you need is not good public relations.
Thus, Janice Haber, OFS Fitel's vice president of systems engineering and market development, told attendees at the OFC meeting that carriers seem to be managing their capital expenditures to meet Wall Street expectations and not necessarily to meet their network requirements. Nelson agrees.
"They don't want to make a signal to Wall Street that they have to deploy new networks or deploy new equipment because Wall Street right now believes they're over-capacitized-incorrectly, since they're not, they're under-capacitized-so any signal that they're buying stuff is [going to be seen as] irrational," he says.
Even with a healthy supply of fiber capacity, Broadwing's DeLong acknowledges that the current economic environment has changed his thinking when it comes to capacity expansion. "Once upon a time, it would have been a very bad thing to ever run out of customer access points on an edge router. In the current economic environment, it is a very good thing to run out of edge customer access ports. Therefore, my engineering team is challenged to essentially enable just-in-time delivery of incremental capacity expansion," he explains. "I have incentive to get much closer to exhaust than I ever was before. However, at the same time, I'm also extremely motivated to be able to increment capacity much quicker than ever before because I'm taking the risk of running out."
Carriers across the United States also face the risk of running out of capacity if bandwidth demand continues to grow at 70% this year. Should that occur, "bandwidth brownouts" might do what current industry data seemingly cannot-convince Wall Street that the investment pendulum must swing again.