|
Many people wonder
about our network here at the Exploratorium, so we've put together a
quick tour:
Our
internal network has recently been rebuilt using equipment generously
donated by 3Com Corporation.
We now have a high-speed ATM fiber backbone network. Whoa! What does
all that mean? Let's take each term separately.
A quick
word about ATM. First, in our case this does not stand for "Automated
Teller Machine." It stands for "Asynchronous Transfer Mode."
This is just the "language" or "protocol" that is
used to send data across the network. It's crucial that all ends of
the network agree on HOW the data is sent (the format, timing, etc.)
before anything can happen at all. ATM is a very efficient method that
allows many different types of data (computer, voice, video) to efficiently
share the same wire or, in our case, optical fiber.
We need to distribute data to hundreds of computers
within the Exploratorium. Most of these machines are
on people's desks. Some of them are used in our store
and admissions desks. A few are in exhibits, and several
are used to supply Internet services like our e-mail
and the Web page you're reading right now. That means
there's a lot of data flowing around the Exploratorium.
Each individual computer may not receive or produce
much data, but collectively quite a few bits are flying
back and forth. If you build a city and want to supply
all the houses with water, you install a few very
large pipes, or water mains. Each house taps into
the water main with a smaller pipe. The water main
carries the collective water for the city. Likewise,
we have a "data main" called a "backbone"
that carries the bulk of the data at a high speed.
Each computer taps into this backbone with a smaller
data "pipe."
Most
data flows through copper wires. If you want really
high speed (many bits per second), another good way
to go is with optical cable. With this method, the
electrical pulses are turned into flashes of light
that are fed into the end of a "wire" made
of glass. This wire is called a "fiber optic."
The fiber optic is designed to allow the flashes of
light to travel great distances without (much) absorption.
When the flashes of light reach their destination
at the far end, they are turned back into electrical
signals. The fiber is our big "data pipe"
(even though it's actual diameter is about the same
as a human hair!). We run this fiber between several
distribution points within the museum. At each of
these distribution points, the optical signal is converted
back to an electrical signal and we use standard copper
wire to connect to the individual computers. The map
(above ) shows the fiber and distribution points (called
"distribution frames") within the Exploratorium.
The MDF is the "Main Distribution Frame"
where all the data converges, and the IDFs are "Intermediate
Distribution Frames" where data goes out to the
various machines.
The distribution
frames are where everyone connects. How do we connect hundreds of computers
to the same optical fiber? We use two devices, one called a "hub" and
another called a "switch."
A
hub is a simple device that will allow many
people to attach their computers together. The hub
makes a connection to all of the machines simultaneously.
For those old enough to remember, this is like having
a telephone "party line," where everybody
in the building had to share the same telephone line.
If someone in apartment 3 was using the phone and
another person in apartment 7 picked up their phone,
they would hear apartment 3's conversation and would
have to wait until 3 was finished before placing a
call. This is efficient usage of a single wire, but
not very convenient for everyone. The hub acts the
same way. Only one device can talk at a time and the
others wait. If two start talking at the same time,
they detect this and each computer waits a random
amount of time and tries again. If your data traffic
is very high, many data collisions take place and
everything slows down. This is where the switches
come into play.
The
switch looks almost identical to a hub, but
it works very differently. The switch works more like
our current phone system. If a computer on one of
its ports needs to send data to a computer on another
port, the switch makes a connection between these
two computers only. None of the other computers plugged
into the other ports on the switch "hear"
anything at all. The computers on the other ports
can also make their own connections because the switch
can allow many "conversations" to take place
simultaneously. The computers don't have to share
a common data path -- each gets its own connection
on demand. This is just like dialing a friend on the
telephone. You get a private wire between both of
you. (Actually, the computer at the phone company
that does this for you is called a "switch.")
As you might guess, switches are much more expensive
than hubs (by a factor of about 10!). It would not
be economical to connect every computer to its own
port on a switch. The compromise is to connect several
hubs to a switch, each hub with its own port. This
way several machines must "share" a hub,
but each hub gets an individual line or port on the
switch. If a machine on one port of a hub talks to
another machine on the same hub, the rest of the network
never sees that traffic. It is therefore wise to intelligently
plan which machines plug into which hubs.
The
switches at the IDFs are hooked together by fiber-optic
cable at the MDF with a special switch capable of
handling multiple fibers all transmitting at very
high data rates. All the fibers converge here and
the data is switched appropriately.
|