Now that last year's financial bubble has burst in the optical communications space, those who were most embarrassed by their excesses have inevitably sought to assess blame for their predicament. The analytical community serves as a popular scapegoat. Capacity demands were widely inflated, say the revisionists, and the analytical community failed to keep a sharp eye on the difference between industry optimism and reality.
To show the depths of this betrayal, many critics have cited a report issued in March by Merrill Lynch that estimates that, on average, "only" 3% of the currently available capacity in the United States is used to carry active traffic. With such a seemingly paltry level of capacity utilization, the spin goes, it's no wonder that optical stocks are tanking on Wall Street. Surely, carriers aren't going to buy additional optical-networking equipment until that 3% increases 20 or 30 times, are they?
Skeptical of that 3% figure, I called the folks at Merrill Lynch and spoke with a pair of analysts, Michael Ching and Simon Leopold, to discover the assumptions behind the estimate. The two tell me that they believe that the 3% number is reasonably accurate. However, many of the conclusions that have been drawn from that figure are not.
To discover network utilization trends, Merrill Lynch researchers used estimates from the Federal Communications Commission, plus input from carriers and market research reports, to determine how many active lines are out there, what kind of traffic they're carrying, how long connections last, and how much bandwidth those connections require. Space doesn't permit a full explanation of the calculations. However, the math works out to an average of 3% for the amount of capacity being used to carry active traffic, and Leopold and Ching stand behind the figure.
What they don't support are the conclusions some have drawn from this data, particularly when it is taken out of context to imply that "the bandwidth glut" has reached monumental proportions that will take many years to correct. First, the two analysts point out, the 3% doesn't include protection bandwidth, which Leopold estimates to be about 40% of capacity on your average fiber network. Also, it doesn't take into account geographic bottlenecks, where bandwidth constraints may actually be acute. Finally, and perhaps most tellingly, 3% is the average figure, which naturally contrasts to peak loads that can be much higher. How much higher, the analysts aren't sure. Anecdotally, Leopold has heard that voice peaks may be three times higher than average, while data could spike as much as 20 times.
When you look at historic usage trends, the current 3% average is indeed low. Average capacity utilization between 1985 and 1990 was about 15%, says Ching. He believes that once the average utilization returns to that 15% level-and maybe before then, now that data represents the majority of network traffic-carriers again will want to add capacity. Merrill Lynch's position is that this will take about two years-well before some of the more gloomy analysts would have you believe.The bottom line is that while some may indeed have been too eager to accept without question rosy prognostications of limitless bandwidth demand last year, one must be careful not to err in the other direction now that the market has taken a turn. In good times and bad, a hard look at the data remains the best path to success.