Part 1: The Nature of the Public
In the next several articles, I
would like to examine the effect of Internet traffic on the public voice
network. Recently, Bell Atlantic has petitioned the Federal Communications
Commission (FCC) to change some rules to lessen this impact on the Telephone
companies and their customers. There are also other reports on some smaller
communities whose customers cannot get dial tone or reach 911 because of
the amount of Internet calls going on during part of the day.
Before we look at how all of that
is possible, we need to understand a little about the nature of the public
voice network. That is the main subject of this article. The next article
in this series will deal with the traffic assumptions of a voice network,
and how the telephone companies use it to engineer their networks. The
third article in the series will describe the Internet traffic, how does
it differ from the regular voice traffic, and how does it impact the voice
network. The last article in the series will speculate on what can be done
to lessen that impact.
The US telephone network is the
largest and most reliable network in the world. It has a long history on
how it got to this point in time. With 90% of the households with a telephone,
the network has become a necessary, and a life line for an increasing number
of people. It is hard for us to imagine how life would be without the telephone.
The telephone network has matured a great deal in the last decade. It now
offers much more than the regular Plain Old Telephone Service (POTS). A
few other types of traffic that it carries include: Paging, Fax, Modem,
and Internet. Some other types of traffic that start to flow on the same
network are: video conferencing, and video entertainment.
How does the telephone network
Conceptually, every telephone in
the world is connected to every other telephone. These connections are
either physically or via some other transmission technologies such as microwave,
radiowave, light, etc. One can imagine 2 persons, each with a tin can connected
by a string talking to one another. The real picture is not much different
than that other than a little more complicated. The complication begins
when we try to connect 10 people together. For this to happen, do we have
9 pieces of string from each person to the other 9 persons? Obviously not,
for several reasons. One is when a person is already on the phone, other
persons cannot reach him/her any more so there is no need for another line
to that person. The second reason is simpler: it is not economical to run
multiple lines to each phone.
To solve this problem, the switchboard
is invented. This is where the single line from each telephone terminates.
If person A wants to talk to person B. A just call the switchboard and
ask to be connected to person B. The operator will logically tie the two
strings together then the model of the 2 tin cans appears again. If too
many people call the operator at the same time, someone will have to wait
to be served. If the wait is too long we may need several operators. As
technologies improve, this operator function is automated and becomes a
sophisticated electronic switch.
Another complication is when the
10 people who are connected to the "operator" live in 2 different
cities. Let's say that 5 lives in Atlanta, and 5 in Los Angeles. Where
should the operator lives? One solution is may be Kansas City. But when
the model grows larger, it is not efficient or economical to have 2 persons,
both live in Atlanta, talking to each other via Kansas City. This is when
we need one set of operators in Atlanta and another one in LA. The people
in the same city can talk to each other using their own local lines. Only
when they wish to talk to people in the other city that the two operators
need to communicate and "tie" the right strings for the communication
to happen. The next question is if there are 10 people in Atlanta and 10
people in LA then how many "strings" do we need between the 2
cities? Obviously, the most would be 10 and the least would be 1. Having
1 connection between cities is not that uncommon in many parts of the world.
In our travel, many of us may have experienced the wait by the phone to
make a call home from our hotel in Europe or Asia until the operator can
get the line out of the country for us to talk. This happens for several
reasons: The long link connecting the cities is often expensive, and there
is not enough information for the planner to know how many links are needed.
We know that not everybody in one
city wants to talk to someone in the other city at the same time. Further
more if a telephone is busy because someone already using it then nobody
can reach that person anymore (we will touch on call waiting, and three
way calling, etc., later on). The problem is reduced to a probability problem:
How many people on the average want to talk between cities at the same
time? What happens if when someone wants to use that link and cannot get
it? Will they go away and try again sometime in the future? Will they re-dial
right away? Will they go to a competitor?
The situation is much more complicated
when we know that there are n cities that need to be connected, the total
number of interconnections can be as much as n(n-1)/2. How many of these
interconnections need to be really connected? How should the call from
Atlanta to LA be routed? One scenario can be as follows: Try to connect
Atlanta to LA direct, if a link exists and is free. If not, try connect
through Kansas City if possible, otherwise try Chicago. If all links are
busy or unavailable then issue a fast busy signal to the caller. Another
method of routing calls is based on what is the best chance for the call
to go through. For example, a call from Atlanta to New York at 9:00 AM
is often best to be completed via Los Angeles. Since at 9:00 AM eastern
time, most people in LA are still in bed. The links to and from the "operator"
or switch in LA are pretty much free and we would have a very good chance
to complete the call. How to route each call is a subject that is being
researched for many years by many people.
As the network becomes larger, it
outgrows the capability of the human operator. The electronic switch can
do better and faster most of what an operator can do. What it lacks is
the intelligence and experience of the operator. Many have attempted to
program these in the switches with limited success. For example, before
the human operator in Atlanta connects a call to LA via Chicago, he/she
would call the operator in LA and ask if the destination phone is free
or not? If it is busy then there is no need to try to connect to any link.
The Atlanta operator can also call the operator in Chicago and ask to see
if there is any free connection from there to LA? If there is not then
he/she may try Kansas City instead. If nothing is available then there
is no point in connecting to any link and ties up that link for no reason.
The telephone network mimics this
forward checking capability of the operator by using a special communication
protocol between the switches. The switches communicate by sending short
messages on a separate specialized Signaling Network using the Signaling
System 7 (SS7) protocol. With these messages, the switches communicate
the status of the links (free or busy), the status of the destination phone,
the best route to take, etc.: The same thing that the operators have done
for a long time.
Technologies have caught up and
provided some service features that an operator can do naturally. Features
such as call waiting, call forwarding, three way calling, caller ID, etc.
have just started to appear in the last 10 years. There are still many
other features that a human operator can do possibly but a machine still
cannot do very efficiently yet. Some examples of these are:
- Two people speak two different
languages and the operator does the translating,
- Routing based on the caller, and
not the calling number. (e.g., If Mr. A calls then route to one number,
but if Mrs. B calls even from the same phone then route to another number)
As technologies continue to improve,
more features will be added. Each of these features will add another dimension
to the traffic and alter the network in new directions. To build the current
telephone network, over a hundred years of traffic measurements have been
collected and studied. Statistics such as length of calls, how often does
people call? what time of day does people call? Is there a difference between
business calling and residential calling? what day of the year is the busiest
day? what time during that day? As new services, and new features are added,
some of the above statistics will change and will affect the total performance
of the network.
In the next article we will discuss
the traffic assumption of the public voice network and how to engineer
various parts of the network based on the traffic measurements. I have
obviously omitted many of the technical details out of the discussion to
make the article readable for everyone. I welcome any technical question
or comment on this article and any other articles in our regular columns.
Please send them to us at firstname.lastname@example.org
and we will discuss them on email@example.com.
Luc T. Nguyen, Ph.D.
For discussion on
this column, join firstname.lastname@example.org
1996 by VACETS and Luc T. Nguyen