A bandwidth glut.
Business Communications Review, Sept 1999 v29 i9 p18(2)
By David Passmore
Summary
The exponential growth of Internet traffic, a pace that seems to be capable of exceeding the current supply of available bandwidth, makes the possibility of bandwidth oversupply an impossible occurrence. But the entry of new Internet service providers, the construction of additional fiber capacity and advances in the field of optical technology promises the availability of a virtually limitless supply of bandwidth and a lowering of the cost.
Full Text
Will network capacity be abundant and cheap? The technology is headed in that direction
One of the more interesting issues these days is how the laws of supply and demand will apply to future network capacity and pricing. With Internet traffic continuing to grow almost exponentially, it's easy to envision a time when bandwidth demand will exceed supply, and thus wide-area bandwidth will remain expensive. On the other hand, a combination of factors - the market entry of new service providers, aggressive fiber construction and spectacular advances in optical technology - promise virtually unlimited amounts of bandwidth and, presumably, an era of low prices.
So which is it to be? Will technological progress and the competition lead to a bandwidth glut, or will the growth in traffic outpace supply?
Tremendous Traffic
Given the almost-insatiable demand for network services, the very idea of a bandwidth glut seems hard to imagine. Even if prices were to fall, given the elasticity of demand, increased usage would more than soak up any excess bandwidth. Whenever the carriers have increased capacity, users have found ways to use it all. While wholesale prices for "fat pipes" in carrier backbones may have fallen recently, the prices paid by business and residential end customers have remained remarkably stable.
Enter the Internet - and IP in general - which rely on a network that is, pardon the expression, "dumb." Which is to say that intelligence resides at the network edge, which lets users or content providers come up with new ways to consume more bandwidth without having to modify the network infrastructure. While it is difficult to predict which applications will become tomorrow's "bandwidth hogs," a good bet is they will be related to entertainment; for example, streaming video at steadily-increasing resolutions (e.g., the equivalent of HDTV channels), real-time interactive videoconferencing and virtual reality gaming.
Future traffic growth also will be fueled by a proliferation of Internet-based application service providers (ASPs). As more enterprises migrate their internal as well as customer-facing applications to server farms operated by ASPs (and as PCs are replaced by Internet appliances), it will be necessary to migrate ever more communications activity to carrier networks, which of course will require more WAN bandwidth.
The increase in multimedia websites will push consumers toward broadband access services, and that means that at the edge of carrier networks, there will be a huge increase in the number of residential users migrating from 56-kbps modems to xDSL and cable modem services, which run at megabit speeds. According to Communications Industry Researchers (www.cir-inc.com), the number of North American households that will sign up for Internet access at 1.5 Mbps or higher is expected to grow from 1.6 million today to 31.7 million by 2003.
Multimedia plays a key role in spurting this growth. A few years ago, the average Web file transfer was about 4 kilobytes; today it's 15 times that amount, rendering traditional dial-up inadequate. Even with contention for backbone bandwidth in DSLAMs and cable headends, improvements in access link speeds will create at least a 10x bandwidth increase in carder traffic.
Another study says that the demand for bandwidth to provide services like voice- and video-over-IP is outpacing supply. The Multimedia Research Group (www.mrgco.com) found that a "bandwidth problem" is likely to persist despite the steps being taken by carders like AT&T and MCI WorldCom to upgrade capacity. Multimedia Research Group concludes, "The bad news is that bandwidth demand continues to grow by some estimates at least three times as fast as supply."
Getting Enough Fiber
Against the backdrop of the anticipation of soaring demand for bandwidth, network service providers are racing to increase the capacity of their fiber backbone networks. The race is driven mostly by the upstart carriers. For example:
* Level 3 (www.leve13.com) expects to finish a 16,000-mile network during 1Q01.
* IXC (www.ixc-comm.com) has completed 12,900 miles and expects to have 16,400 miles of fiber in the ground by the end of the year. (Note: IXC is being acquired by Cincinnati Bell.)
* Qwest (www.qwest.com) just completed construction of 18,400 miles of fiber, and that doesn't include the facilities it will acquire when - and if - its merger with US West goes through.
* Williams (www.williams.com) hopes to complete a 32,000-mile fiber network next year.
* Frontier (www.frontiercorp.com), which will be acquired by Global Crossing, expects to finish its 20,000-mile network by year-end.
* Enron (www.enron.com - a natural gas utility) plans its own nationwide fiber network.
* Global Crossing (www.globalcrossing.com) has introduced competition, lower prices and increased capacity on undersea fiber routes.
Meanwhile, the incumbent long-distance providers continue to augment their fiber networks; AT&T already has 38,700 fiber miles, MCI WorldCom has 44,800 and Sprint has 23,500 miles. Most of the fiber has not yet been "lit," and in the case of the new market entrants, typically less than 5 percent of their fiber is currently in use.
Despite the small percentage of fiber that's actually carrying bits, fiber's carrying capacity is exploding as a result of dense wave-division multiplexing (DWDM). Today's DWDM can already enable a single fiber to support 80 Gbps (40 OC48 channels, each at 2.5 Gbps) or 160 Gbps (16 OC-192 channels, each at 10 Gbps). By 2003, a single fiber may be able to support about 500 OC768 channels (each 40 Gbps), for a total of 20 Tbps - terabits/sec.
DWDM is increasing the bandwidth for fiber at a much faster rate than Moore's Law propels increases in CPU capacity. If you multiply DWDM's effect by the amount of fiber being installed, the amount of bandwidth that could become available is mind-boggling.
Supply vs. Demand
So far, this discussion has focused on bandwidth supply. To get a handle on the demand side, consider how traffic patterns are changing.
Most future traffic is expected to be asymmetrical: a great deal moving in the server-to-client direction, with very little originating from the client. This is consistent both with today's Web traffic patterns, and streaming audio or video. Interactive voice and video (not streaming) traffic is symmetrical, but is expected to be a smaller percentage of total traffic. Access to network-based applications (e.g., those hosted by ASPs) also will result in highly-asymmetrical traffic.
The data rate limit for a video connection is based on the number of pixels being displayed, and the rate at which pixels are updated. (Vector graphics or HTML/character screens require far less bandwidth, as does audio or voice.) At present, the highest display bandwidth requirement being envisioned is HDTV: 1,080x1,920 pixels at 30 frames/second. A single compressed HDTV channel requires 19.2 Mbps.
One fiber with 500 wavelengths of OC-768 (expected by around 2003) can transmit more than 1 million simultaneous HDTV streams (or the entire Library of Congress in half a second). It is not unusual for a single carrier cable to contain at least 100 optical fibers, so a cable could theoretically carry 100 million high-definition TV channels (although there'd probably still be nothing on!). based on this very simple analysis, it's hard to see how there will be much of a bandwidth shortage.
Pricing Trends
Need additional proof that a bandwidth glut is coming? Consider how prices have started to fall. As just one example of this trend, GTE Internet-working (www.bbn.com) recently reduced its dial access Internet pricing by 30 percent in response to competition from higher-speed cable modems and xDSL services.
Even more importantly, the prices carriers pay when they sell bandwidth to one another - i.e., the wholesale spot market for bandwidth - are coming down. While carriers have traditionally negotiated long-term contracts, pricing pressures and new competition are creating an opportunity for online bandwidth brokerages.
These brokerages (for example, RateXchange - www.ratexchange.com, which is about to be acquired by NetAmerica.com) let carders anonymously enter into short-term contracts for bandwidth. The carders wind up undercutting their own advertised rates.
The existence of these exchanges demonstrates the extent to which bandwidth has become a pure commodity. Enron takes this to its natural conclusion by its efforts to establish a bandwidth futures market, similar to the oil and gas industry.
Within the past year, brokerage pricing shows how the market has transitioned from scarcity to abundance. The RateXchange average price for transmitting a megabyte of data over domestic or international lines has fallen by 25 percent. Similarly, the cost per minute for telephony has fallen about 30 percent. Such falling prices mean that carders have to carry more traffic just to maintain the same revenues.
So how come most enterprises aren't enjoying falling prices for bandwidth? Well, enterprises normally don't lease just long-haul fiber bandwidth - they contract for end-to-end network services, and that has to include local access.
And there's the problem. It's going to take a while for customers to gain cheap and plentiful access to long-haul bandwidth. That, in turn, will make the situation for long-haul carders even worse; the access bottleneck will limit demand for interexchange bandwidth, which will only accelerate price drops among these carriers.
Ultimately, the combination of new carders, new technologies and regulatory relief should result in access bandwidth becoming cheaper and more plentiful. As that happens, customers will be able to take advantage of inexpensive, high-capacity interexchange bandwidth. While a bandwidth glut is not certain, enterprise managers should begin thinking about how they might exploit such a positive development.
David Passmore is research director and founder of NetReference, Inc. (Sterling, VA), a consulting firm specializing in networking strategies for vendors and end users. He previously ran Gartner Group's networking services, was a partner with Ernst & Young and was a principal and cofounder of Network Strategies. He can be reached at passmore@netreference.com.
Full Text COPYRIGHT 1999 BCR Enterprises, Inc. |
