What Will Be the Bottlenecks ?
Bandwidth Gluttony
As the Internet backbone rapidly builds out, congestion centers on local networks.
By Rolf De Vegt and Elizabeth Steels
thestandard.com
Long-haul fiber-optic networks are being built at an unprecedented rate. Transatlantic bandwidth on the Internet backbone is expected to reach 10 terabits per second by 2001. New European carriers, such as U.K.-based consortium I-21, have promised "petabit" (1,000 terabits per second) networks over the next couple of years.
Will this "fiber frenzy" result in a capacity glut, or will demand surge to fill these fat pipes?
We predict that bottlenecks in the access portion of the network (the "last mile"), along with a shortage of high-bandwidth applications, will limit near-term demand. The resulting capacity glut will likely persist well into 2005.
IF THEY BUILD IT ...
To meet the demands of the boundless Internet, many network operators such as Teleglobe (TGO) and Global Crossing (GBLX) are building on a global scale. Other carriers are developing pan-European, transatlantic and North American networks.
This tremendous amount of bandwidth is the result not only of the number of fibers being laid, but also of switching and transmission advances in the form of Dense Wave Division Multiplexing. We are moving from single wavelengths to 160 wavelengths multiplexed onto single fibers.
Many new networks being laid run between 96 and 144 fiber pairs per conduit, so the total bandwidth potential available on a single network could be sufficient to carry current U.S. and European traffic at least 20 times over. Surprisingly, the scale of the potential supply overbuild is greater in Europe than in the United States. This is due to the larger number of new market entrants in Europe, with at least 18 new players investing in networks that cover Europe's core economies, and to the higher fiber count of many new European networks, such as I-21 with 192 fiber pairs.
... WILL THEY COME?
These networks are being built with the expectation that as capacity becomes available, demand will emerge as users adopt newly available higher bandwidth applications and operate existing applications at higher speeds. This optimism is based on current trends in Internet traffic, which is estimated to double every 100 days. Internet penetration is expected to grow to over 65 percent in the United States and Europe by 2004.
Growth in Internet demand is driven by several factors, including the emergence and demand for bandwidth-hungry applications such as streaming media. Dataquest anticipates that the number of U.S. firms that use streaming media on Web sites and intranets will increase from 9 percent to 30 percent within a year.
Consumer-requested dialup videostreams, meanwhile, will increase from 171 million in 1999 to 1 billion in 2005, according to Webcast Track. This is powering the increase in available backbone bandwidth and fueling the growth of streaming-media companies such as Real Networks.
However, the degree to which streaming media stimulates demand for bandwidth depends on several things. First, the rate at which users will adopt streamed media is open to question. Video conferencing, for example, has been "about to take off" for over a decade, and has yet to deliver the desired quality at a price that spurs widespread adoption.
Second, caching is likely to affect the bandwidth demand that streamed-media applications put on the backbone network. Caching technologies, such as those developed by Akamai, store frequently accessed Web pages and content on servers at the edge of the network, thereby reducing pressure on the backbone. The Internet Research Group estimates that caching technology could reduce demand for backbone bandwidth by as much as 35 percent. |
