20.2
INTRODUCTION TO ATM
513
By contrast, most LAN protocols are connectionless. This means that LAN nodes
simply transmit traffic when they need to, without first establishing a specific connection
or route with the destination node.
In that ATM uses a connection-oriented protocol, cells are allocated only when the
originating end user requests a connection. They are allocated from and idle-cell pool.
This allows ATM to efficiently support a network's aggregate demand by allocating cell
capacity on demand based on immediate user need. Indeed it is this concept that lies at the
heart of the word asynchronous (as in asynchronous transfer mode). An analogy would
help. Let's say that New York City is connected to Washington, DC, by a pair of railroad
tracks for passenger trains headed south and another pair of tracks for passenger trains
headed north. On these two pair of tracks we'd like to accommodate everybody we can
when they'd like to ride. The optimum for reach this goal is to have a continuous train
of coupled passenger cars. As the trains enters Union Station, it disgorges its passengers
and connects around directly for the northward run to Pennsylvania Station and loading
people bound for New York. Passenger cars are identical in size, and each has the same
number of identical seats.
Of course, at 2
A.M.
the train will have very few passengers and many empty seats.
Probably from 7
A.M.
to 9
A.M.
the train will be full, no standees allowed, so we'll
have to hold potential riders in the waiting room. They'll ride later; those few who try
to be standees will be bumped. Others might seek alternate transportation to Washing-
ton, DC.
Here we see that the railroad tracks are the transmission medium. Each passenger car
is a SONET/SDH frame. The seats in each car are our ATM cells. Each seat can handle a
person no bigger than 53 units. Because of critical weight distribution, if a person is not 53
units in size/weight, we'll stick some bricks in the seat to bring the size/weight to 53 units
exactly. Those bricks are removed at the destination. All kinds of people ride the train
because America is culturally diverse, analogous to the fact that ATM handles all forms
of traffic. The empty seats represent idle or unassigned cells. The header information
is analogous to the passengers' tickets. Keep in mind that the train can only fill to its
maximum capacity of seats. We can imagine the SONET/SDH frame as being full of cells
in the payload, some cells busy and some idle/unassigned. At the peak traffic period, all
cells will be busy, and some traffic (passengers) may have to be turned away.
We can go even further with this analogy. Both Washington, DC, and New York
City attract large groups of tourists, and other groups travel to business meetings or
conventions. A tour group has a chief tour guide in the lead seat (cell) and an assistant
guide in the last seat (cell). There may be so many in the group that they extend into
a second car or may just intermingle with other passengers on the train. The tour guide
and assistant tour guide keep an exact count of people on the tour. The lead guide wears
a badge that says BOM, all tour members wear badges that say COM, and the assistant
tour guide wears a badge that says EOM. Each group has a unique MID (message ID).
We also see that service is connection-oriented (Washington, DC, to New York City).
Asynchronous means that we can keep filling seats on the train until we reach its
maximum capacity. If we look up the word, it means nonperiodic whereas the familiar
E1/DS1 are periodic (i.e., synchronous). One point that seems to get lost in the literature
is that the train has a maximum capacity. It is common to read that ATM provides
"bandwidth" on demand. The underlying transmission medium has a fixed bandwidth. We
believe the statement is supposed to mean "cell rate capacity" on demand. This is true of
course until we reach the maximum capacity. For example, if the underlying transmission
format is SONET STS-1, it can provide 86
× 9 octets × 8000/(53) per second or about
116,000 cells per second of payload (see Chapter 19).