It’s time to expand by one the number of quests to solve the great mysteries of the universe. Along with figuring out the origin of those weird-looking statues on Easter Island, unraveling the mysteries surrounding the curse of King Tut and tracking the dryer-to-ether journey of the common sock, scientists, analysts and marketing experts are now looking to resolve the looming conundrum of the 21st century: Is Internet traffic increasing or slowing down?
While a precise measurement of Internet traffic was never a major concern during the salad days of network buildouts and steady double-digit revenue growth for equipment makers, interest in the topic has gained significantly in the past year or so. Not surprisingly, that’s about the time the bottom fell out of the telecommunications industry.
In a scramble to create a correlation between Internet-traffic growth rates and the current telecommunications downturn, Internet and financial experts have been trying to get a handle on past and future Internet growth. The major justification behind the new research, of course, is to both provide an explanation for current conditions and offer some sort of guidance on the future spending habits of major carriers.
The major obstacles to realizing this objective, however, is the existence of conflicting measurements and theories regarding the relationship between Internet growth rates and the current cutback in capital expenditures by carriers.
Essentially, three theories exist about the way the Internet has been growing since 1996, the year the National Science Foundation stopped providing statistics on traffic growth.
One theory, which was recently offered by Lawrence Roberts, one of the pioneers of the Internet and the current founder and chief technical officer at startup Caspian Networks, describes a robust Internet that is currently expanding in terms of traffic faster now than it has is previous years.
“What we concluded is that Internet traffic growth has increased in 2000 and is increasing into the first half of this year,” says Roberts.
Roberts’ findings, which are based on analysis of traffic crossing the high-speed backbone trunks of 19 Tier 1 service providers, indicate that Internet traffic is currently multiplying at a rate of four times annually. The current growth rate, says Roberts, is actually faster than that before April 2000, which was roughly 2.8 times a year.
The findings by Roberts, one of the designers of the ARPANET network, which later evolved into the Internet, offer a stark contrast to a recently published report by analysts at McKinsey & Co. and JP Morgan suggesting that traffic growth had started to slow and would decline to a 60 percent growth rate by 2005. It also conflicts with recent statements by industry leaders, such as Nortel Networks CEO John Roth and industry research firms, such as Ryan Hankin and Kent.
Andrew Odlyzko, a longtime research at AT&T labs and now the director of the Digital Technology Center at the University of Minnesota, proposed a third theory a couple of years ago. In a paper he co-authored called “The Growth of the Internet,” Odlyzko argued that Internet traffic went through a tremendous growth spurt in 1995 and 1996, at about the time the World Wide Web was introduced and e-mail became a standard form of communication. Since then, however, he maintains that Internet growth has stabilized at about 100 percent annually. That number, he says, is based on year-to-year growth rates between 70 and 150 percent.
Odlyzko says his research, which was not, as was Roberts’, done with the assistance of the world’s major carriers, was motivated by his outrage over frequently quoted estimates that Internet traffic was multiplying by eight or more times a year.
“It was unbelievable,” he says, “that people were not asking about the source of these claims.”
In the late 1990s, estimates of traffic doubling every three to four months were routinely cited by industry analysts and telecom-equipment vendors. Odlyzko says those numbers were exaggerations and are largely responsible for the current slowdown in telecommunications spending.
“My view is that what happened is that people were basing their buildout plans on false assumptions,” says Odlyzko. “Then they discovered that the emperor had no clothes.”
Although service providers eventually put a plug on the money gushing out of their coffers, how could they have been fooled for so long? Even if analysts and equipment makers were telling whoppers about actual growth rates, wouldn’t carriers have an honest assessment of the traffic on their backbones and buy only the equipment they would need?
The problem with that logic, say Odlyzko, is that although carriers have a good idea of the traffic on their own networks, they have to rely on the same anecdotal information as everyone else to figure out what’s going on outside their own little worlds. An additional factor is that carriers believed they had nothing to lose by building out their networks at such an incredible pace. The prevailing sentiment of the time was, “build it and they will come.” And build carriers did, believing that traffic would eventually grow — regardless of the exact rate — to fill the pipe.
The Roberts report is in agreement with Odlyzko in that both believe that the rate of growth of the Internet is not in decline. The major difference, of course, is that Roberts believes the Internet is growing at a much faster clip. In terms of methodology, Roberts’ findings are based on data acquired from the top 19 carriers, statistic information that Odlyzko did not have access to.
A potential hole in Roberts’ theory that Internet growth has been robust over the past few years and is now accelerating is the current market condition. If demand is so great, why have carriers put the brakes on equipment purchases?
Robert’s explanation is that carriers, desperate to keep ahead of demand and flush with fresh financing, purchased more equipment than current demand required. In addition to buying a little overhead, carriers could buy a little time between major expansions because they were suddenly applying more stringent deployment practices in the face of slowing revenue growth.
“They are using equipment much more efficiently,” says Roberts.
Confirming that assessment is Mike McAndrews, director of Internet services at Williams Communications. “This is a time to be smart and get more bang for the buck,” says McAndrews. “People tend to get free and easy when the money is flowing.”
The McKinsey and JP Morgan report, issued in May of this year, has been cited as evidence that carriers are buying less infrastructure gear because demand has slackened. Proponents of a slowdown theory say that the decrease in bandwidth demand is responsible for the reduction in equipment purchases because carriers do not need to build out networks as rapidly as in the past.
Of course, the biggest impact of any of these theories is the potential influence on carriers’ buying plans. If the slowdown scenario is accurate, carriers could conceivably stretch out their current infrastructures to handle capacity increases for the next 18 months to two years. If Roberts is right, and traffic is doubling every six months, carriers could be forced to restart their next-generation networking buildouts within six months.
“You’ll see carriers buying equipment heavily by the beginning of next year,” says Roberts.
As a sort of a byproduct of Roberts’ study, industry observers now have a better idea of the competitive landscape in the Internet-backbone market. According to the Caspian research, the largest carrier of Internet traffic only handles about 14 percent of the entire market. And the top four carriers, which make up about 50 percent of overall traffic, are neck and neck in terms of traffic handling. Roberts’ research was done under non-disclosure, and none of the 19 carriers that participated in the survey was identified. Even without a way to identify individual carriers, however, the information in the survey has dispelled one widely held industry assumption: that WorldCom was responsible for more than 30 percent of Internet traffic.
The fact no one carrier has a commanding share of the market will spark competition, playing to the advantage of equipment providers, says Roberts.
“None of these people can afford to let service deteriorate too much,” says Roberts, adding that unless a carrier keeps pace with competitors in terms of next-generation technology and capacity thresholds, it risks losing its business to a rival.
Somewhere in the middle is Odlyzko’s theory, in terms of the timing of a spike in carrier spending.
“At some stage, people will start buying again,” he says. “I don’t know if it will happen in six months. At some point, however, people will start buying for no other reason than technology is advancing.”
While all three theories deal with raw traffic numbers, most fall short of providing information that will shed light on the major problem facing service providers: the unhealthy relationship between profits and capital costs. One of the most crippling artifacts of the massive buildout of IP networks over the past couple of years is the lack of a proportional boost in revenue. Unless service providers can do a better job of earning money off of IP services by offering class of services and service level agreements, buying more equipment will continue to lead to bankruptcy.
The good news for carriers is that advances in technology should help reduce capital and operating expenses. For example, the deployment of optical switches and centralized control-plane software, such as the Generalized MPLS standard, will let networks be provisioned more efficiently, says Derek Hodovance, vice president of IP architecture at Qwest Communications.
Currently, most carriers tend to build networks to run at 40 percent utilization rates. “The reason for that is you have to stay below 50 percent for fail-over scenarios,” says Hodovance. A GMPLS-controlled network with optical switches, he adds, would make it possible to use some of the bandwidth that is currently stranded for backup purposes.
The upshot of all this is that carriers will eventually be able to stretch their budgets further, utilizing more bandwidth without buying additional equipment.
Although it was difficult to find anyone to challenge Robert’s methodology or his qualifications, several analysts, who asked not to be quoted, pointed out that the findings are in the best interest of Caspian. Like most startups, Caspian is counting on network growth to fuel future revenues.
Another element that was largely missing from the reports is a characterization of the traffic that will be fueling Internet growth. Roberts, however, said that the growth of corporate traffic, fueled by the recent requirement to move business online, is the largest contributor to Internet growth.