The Internet simply wouldn't function if it wasn't for the various
"backbone" networks. Danny Bradbury takes a look at a marvel of
modern communications engineering
In the five years since the Web took off, the Internet has exploded
at an incredible rate. It is easy to forget how packets of
information get from one place to another across it, and to take
for granted the technology that makes it happen. None of it would
be possible unless the Internet "backbone" had developed.
The backbone to the Internet evolved after the formation of the
root network, called Arpanet, in the late 1960s. Other networks
developed, such as the National Science Foun-dation's NSFnet, which
was aimed at connecting academic institutions.
In 1990, the US govern-ment stopped civilian traffic running
over Arpanet, and routed it across NSFnet instead, upgrading the
speed of the backbone to 45mbps. As late as 1995, NSFNnet was
privatised, and that gave the backbone companies the chance to get
in on the act. These days, the Government, academic institutions
and some commercial suppliers are working on applications and
equipment that will drive a second generation Internet at much
higher speeds. The National Science Found-ation's Very High-Speed
Backbone Network Service (VBNS), is being used to help support this
activity.
The various private backbones in the UK and elsewhere connect
together in large facilities called Internet exchanges.
In the UK, the most popular of these is Linx, the London
Internet Exchange. Based in the Docklands area of East London, this
facility connects together multiple backbones onto a high-speed
link out of the UK, into Europe and through to the US. In the US
market, there are similar facilities, although these have often
been owned by single large backbone providers, in contrast to the
UK, where Internet exchanges are non-profit organisations jointly
owned by service providers.
The US WorldCom, for example, was given a contract by the
National Science Foundation in 1993 to handle the Internet
exchanges to its backbone, which were called Network Access
Points.
So, how do your messages get onto the backbone? Generally, they
will go via the local loop growth (either dial-up phone, ISDN,
leased line or, if you're really lucky and are in BT's good books,
ADSL), to a local point of presence (POP). From the POP, your
traffic will travel along the backbone to Linx, or another Internet
exchange, from where it will jump onto an international
backbone.
If you are using a very large ISP, then the chances are that you
will get a better service from the backbone. Large carriers, such
as BT, have their own backbone already laid. Using a company like
this as an ISP guarantees that your traffic is as close to the
backbone as possible at an early stage in its journey across the
Internet. More to the point, the traffic has better access to that
backbone.
The danger of partnering with a smaller provider is that they
will be renting backbone bandwidth from a larger backbone owner, so
their bandwidth on the backbone will be more restricted. This can
present problems if the company grows more quickly than expected
and ends up with congested traffic that can't all fit into the
available bandwidth.
For a company running an e-commerce site, the throughput to the
server becomes very important because it affects the number of
customers that can be served at any one time.
One way to guarantee a better throughput is to circumvent the
local loop altogether and collocate the server at the Internet
exchange itself, thereby placing it right on the backbone.
Typically, this will give access to backbone speeds in excess of
1gbps.
One of the biggest problems for backbone providers is that the
world is moving towards a quality of service (QS) model. In this
scenario, Internet packets are differentiated according to their
importance, and importance can be based on different issues such as
the type of data being sent, or how much the customer is paying.
If, for example, you are sending video down the line, you are
likely to want your traffic to travel as quickly as possible along
the backbone, so that the video arrives smoothly at the other
end.
Because Internet data travels in packets, this has been
difficult to achieve in the past. Instead, packets have been
divided up, routed along different paths in the network and
reassembled at the other end as they arrive, leading to scrappy
delivery of data. Moving towards a quality of service model,
backbone providers will tag the packets as they enter the edge of
the network, subsequently routing packets through the network with
different priorities based on those tags.
This is all very well in theory, and is feasible when dealing
with one network, but what happens when one backbone provider has
to pass data to another, so that it can reach a destination on the
other backbone? It is not guaranteed that one backbone provider
will use the same quality of service mechanisms as another.
Consequently, several technology standards are in development at
the Internet Engineering Task Force (IETF). Diffserv, for example,
is designed to enable packets to be prioritised, as is
Multi-Protocol Label Switching (MPLS), and hopefully these will be
adhered to by the different backbone providers when they exchange
packets in the future.
In the meantime, QS along the backbone is not guaranteed,
especially when dealing with multiple backbone providers.
Be sure of one thing, however - when sending your packets down
the line, they embark upon an incredible journey before reaching
their final destination.
Terrific traffic
Like the contents of handbags, wallets and most car glove
compartments, the traffic on the Internet always expands to fill
the available space. While backbone bandwidth may be increasing
exponentially, market research firm Datamonitor predicted in early
1999 that the volume of Internet traffic would surpass the volume
of voice traffic this year. It also found that IP traffic had been
rising at roughly 1,000% a year, compared to a growth of under 10%
in the public switched telephone network (PSTN) area.
Predicted growth of Internet traffic
| | 1997 | 1998 | 1999 | 2000 | 2001 | 2002 |
| Internet
traffic | 3.5 | 18.4 | 94.2 | 451 | 1,773 | 6,170 |
| PSTN
traffic | 280 | 310 | 330 | 350 | 350 | 320 |
Source: Datamonitor