A good article about broadband on raging Bull
Who Will Win the Broadband
Race?
January 5, 1999 - 2:02 PM
By Roderick Beck
Although the Internet is flourishing with strong growth in traffic,
subscribers, and electronic commerce, the lack of speed
threatens to limit its potential. Without faster throughput, the
Internet will never be able to offer a television-like experience or
support services like Internet telephony, streaming video, and
videoconferencing.
Indeed, for the most part, IP telephony and videoconferencing
run over private networks. Let us now take a better look at
exactly why the Internet is slow and how some new
technologies may help it realize its potential as the true digital
highway.
Why Is the Net So Damn Slow?
The speed of a network depends largely on its capacity
utilization. A fast network is one with ample capacity to meet
peak demand. Unfortunately, the Internet is congested -- peak
demand far exceeds capacity. To understand the problem, one
needs some insight into the unique nature of the Net.
The traditional definition of a network is that it uses compatible
hardware tied together by a protocol or a language that allows
the different parts of the network to work together. For
example, consider a phone call from New York to Los Angeles.
A protocol, SS7, enables the New York long distance switch to
contact the Los Angeles switch and create a circuit connecting
the two parties.
The trouble with protocols is that they are typically hardwired or
embedded right into the network hardware and hence unique to
it. This makes it difficult to transfer information between
networks using different hardware and protocols. The fathers of
the Internet solved this problem by developing the Internet
Protocol (IP), a universal language, which allows any computer
network to be seamlessly linked to any other network.
So what is the Internet? The Net is hundreds of thousands of
physically distinct networks united by the Internet Protocol into
a single decentralized network. IP is the linchpin. The Internet
includes the local area networks (LANs) of corporations and
universities, the Internet access facilities in local calling areas,
and the Internet backbones.
The multitude of networks which help make up the Net explains
why it is so often a lackluster performer: there is no central
management to ensure each building block has enough
capacity to ensure high speed. A network is only as fast as its
slowest link. The Internet backbones are national and global IP
data highways owned by different providers such as Worldcom,
Sprint, and a host of others.
Is The Net a Victim of Under Investment?
Some of these providers are more capable than others. While
competition may eventually force less competent providers to
improve, there is a more serious problem. Because Web sites
and Internet service providers use different backbone providers,
the backbones themselves are linked to each other via network
access points (NAPs) to ensure universal coverage. For
example, if the customer of an ISP using Worldcom's
(WCOM) backbone wants to download a Yahoo! (YHOO)
website connected to Sprint's (FON) backbone, then the two
backbones must be connected to exchange traffic.
Providers do not reap the full benefits of investing in the NAPs
because additional capacity benefits everybody, including
one's competitors. Hence, under-investment is the norm and
the NAPs suffer from congestion and constant traffic jams.
However, the point is more general: it is a theorem in
economics that the optimal investment occurs only if the
investor receives all the benefit from it. If some of the benefit
goes to third parties, then under-investment is the result. This
perhaps sheds light on the relatively poor performance of the
public Internet vis-à-vis privately owned phone systems and
data networks. For example, Sprint captures the full benefit of
investing in its own long distance system, but it shares the
benefit of its Internet backbone with other backbone providers
who exchange traffic with it.
Flaws in the Internet Protocol?
There are other obstacles as well. When the Net is overloaded,
the Internet Protocol has no way of giving priority to traffic
based on how sensitive it is to delay. It is first come, first
service. Applications like voice, real time fax, and streaming
video are ruined by congestion. Finally, it is questionable
whether Internet pricing provides the right incentives for network
investment.
Economists have pointed out that economic efficiency in the
case of the Internet requires usage-based pricing, yet Internet
service charges are based solely on the size of the connection
to the Internet, not on usage. When Sprint Canada recently
tried all-you-can-eat long distance calling for a fixed monthly
fee, their network almost collapsed under the resulting surge in
traffic and the promotion was quickly modified.
Any network, whether gas, electrical power, phone service,
water or the Internet will be plagued by shortages in the
absence of usage-based pricing because suppliers do not have
adequate incentive to invest and users will tend to overuse it.
Will the Local Phone System Hold Back Broadband?
Another key problem in realizing the Broadband Future is the
limited capacity of the local phone system. Most consumers
and business users connect to the Internet via standard copper
wire. This wire can only carry 64 kilobits per second -- fine for
voice, but too slow for big data transfers or multimedia. Even
large corporations that have fiber connections are not much
better off because their PCs have 56K modems.
The nature of the problem is understandable. The local phone
system was designed for voice, not data. Voice is narrowband,
but data is broadband. The crux of the problem is that we are
using a voice network to carry data. What is important to note
is that it is not clear that eliminating this bottleneck will usher
in a Golden Age of Broadband.
The Net, itself, is congested and the preceding analysis
suggests that no quick fix is on the horizon. Price reform is
very unpopular and much of the Internet's congestion reflects
the fact that it is a public network with no central management. |
