There and back again... how your data travels
on the Net
by Alan Zisman
(c) 1996. First
published in Computer Player, September 13, 1996
When Bilbo Baggins, returned from his adventures, he
called the story
of his adventures (since reprinted as "The Hobbit") "There and Back
Again".
When you're looking at a Web page, on the Internet, perhaps a continent
or two away, the data you're accessing could tell (if data could talk
or
write!) a tale of a similarly involved journey.
How Internet data, whether a Web page or an e-mail
message gets from
here to there is a mysterious-seeming process-but understanding it
helps
to make sense of other mysteries, like who really pays for the Net, and
why your connection sometimes seems so slow.
Let's see what really happens when I, in Vancouver,
using my modem at
home or at work, try to connect up to a Web page... say Time Magazine's
Pathfinder site, in New York City. First, using my modem, I connect to
a local Internet Service Provider... one of over 2,300 such services in
North America. My modem tries to get a connection at its maximum rate,
28,800 bits per second, but depending on the amount of noise on the
phone
lines, will often end up settling for a slower connection.
At the Service Provider, I connect to another modem,
which is connected
to a computer, connected to a local area network-connecting the ISP's
bank
of computers, modems, and phone lines. There's one modem per phone
line-and
sometimes all the lines are in use. If I don't get a busy signal, I'm
connected
to the ISP's Ethernet network connecting all their computers at a rate
of 10 million bits per second (Mbps... about 300 times faster than the
modem connection). I can check my e-mail or Usenet news groups-in both
cases, looking at messages stored on my ISP's network. But to send
mail,
or use the Web, I've got to go out from my ISP-over the Internet
itself.
The Internet is a network of networks. Regardless what
sorts of computers
they are, they transmit data by a common code-TCP/IP, for Transmission
Control Protocol/Internet Protocol. The TCP part describes how data is
broken down into small pieces called packets... each packet is
transmitted
separately, mixed with other data-and sometimes travelling along
different
routes. At the end, it's all reassembled, hopefully in the correct
order.
If a packet is missing or corrupted, TCP will request it again.
The IP part gives every data packet its Internet
address... a 12-digit
number that corresponds to the more human-friendly address that you or
I typed. (I know, Internet addresses like http://www.tcp.com doesn't
seem
that friendly, but try to remember 167.191.045.001 instead). One of the
computers at my ISP is the Domain Name Server-storing a large table
that
translates the Internet address that I typed into its (hopefully)
proper
numerical equivalent.
To get out to the Internet, my ISP rents a dedicated
T1 connection,
running at 1.54 Mbps, connecting to a larger, regional ISP. From there,
my TCP/IP packets go through a router-a sort of dedicated computer that
does nothing but try and figure out the most efficient route to get the
data to its destination. From Vancouver, on its way to New York, the
data
may go via Seattle, or down to Sacramento, travelling along a 45 Mbps
T3
connection owned by MCI. Not only is data being exchanged however...
despite
our sense that the Internet is free, each of these data transmissions
is
being paid for, as the various Service Providers rent their access from
larger companies such as MCI.
While the T3 line sounds massive, much of the
so-called backbone-the
main transmission route, is overloaded. MCI is currently replacing it
with
fibre-optic cable capable of 622 Mbps. Moving on the backbone, my data,
mixed with packets from thousands of other users on hundreds of other
ISPs,
may pass through routers in Denver and Kansas City, on its way to New
York.
There, it passes through a series of progressively smaller Service
Providers,
on its way to its destination, at Time-Warner's Pathfinder Web site.
(From there, the data needs to get back to me-the same process, though
perhaps a different route, in reverse).
Amazingly, this complex process works correctly most
of the time. And
while there are sometimes bottlenecks along the way (gridlock on the
Information
Superhighway, if you'll excuse the cliché), most of the traffic
problems occur at the destination. Sites like Pathfinder are quite
popular,
and if several thousand people are all trying to access it at once, it
can take a while to connect.
The Web is designed to try to minimize this,
however... when I'm reading
that Time Magazine article, I'm not actually connected to their
site-Web
browsers are only actually connected for the few seconds when they are
sending or receiving data. Many people can be reading data from a site,
taking turns connecting, before performance starts to degrade.
You can see the route your data takes, using the
Traceroute utility.
Search for 'traceroute' at a search engine like AltaVista
(http://www.altavista.digital.com/),
to get links to Web sites with this program, which will show you the
journey
your data takes from There and Back Again.