 |
|
 |
 |
Packet Switching and Circuit Switching; An Always On
Connection
Written by T.Farley
"Thanks to more capable electronics for handhelds, communications
companies are scrambling to deploy so called 2.5G (for generation 2.5)
networks more attuned to the world of data. In earlier networks, whether
analog or digital, each call creates a circuit that reserves a channel
between two parties for the entire session. The 2.5G devices are the
first to use Internet-style packet switched networks; they send bursts
of data only when needed. Because these devices don't hog an entire
circuit, they can be "always on."
John Ueland, writing in the article 'Internet Everywhere', from the
September/October issue of MIT''s Technology
Review.
There's much talk about the coming mobile internet, about how people
will have a wireless, always on connection to the web. How will that come
about? In two words, packet switching, a fundamental, elemental change
between how wireless was delivered in the past and how it will be presented
in the future.
Conventional cellular radio and landline telephony use circuit switching.
Ricochet's
wireless modems and wireless services like Cellular Digital Packet Data
or CDPD, by contrast, employ packet switching. Wireless services now developing
such as General Packet Radio Service or GRPS, Bluetooth, and 3G, will
use packet switching as well.
Circuit switching dominates the public switched telephone network or
PSTN. Network resources set up calls over the most efficient route, even
if that means a call to New York from San Francisco, for example, goes
through switching centers in San Diego, Chicago, and Saint Louis But no
matter how convoluted the route, that path or circuit stays the same throughout
the call. It's like having a dedicated railroad track with only one train,
your call, permitted on the track at a time.
Before we go on, let's talk about digital. Voice and data from the local
loop goes digital once it hits the local telephone switch. Traffic between
American telephone offices is nearly all digital, you know, 1s and 0s.
Bits. That includes most circuit switched traffic, like we just discussed.
All these bits get packaged into small groups called packets, frames,
blocks, or cells. T-carrier, SONET, ATM, frame relay, pick your transmission
technology, all traffic gets put into one form of packet or another. But
simply packetizing data does not mean a call is packet switched. (A caveat,
systems like SONET combine elements of transmission and switching in one,
calling them strictly a transmission method is a little too simple, but
enough for our discussion here.)
Packet switching dominates data networks like the internet. A data call
or communication from San Francisco to New York is handled much differently
than circuit switching. With circuit, all packets go directly to the receiver
in an orderly fashion, one after another on a single track. Like the train
we mentioned before, hauling one boxcar after another. With packet switching
routers determine a path for each packet or boxcar on the fly, dynamically,
ordering them about to use any railroad track available to get to the
destination. Other packets from other calls race upon these circuits as
well, making the most use of each track or path, quite unlike the circuit
switched calls that occupy a single single path to the exclusion of all
others.
Upon getting to their destination, the individual packets get put back
into order by a packet assembler. That's because the different routes
practically ensures that packets will arrive at different times. This
approach is acceptable when calling up a web page or downloading a file,
since a tiny delay is hardly noticed. But one notices even the tiniest
delay with voice. This point is really important. Circuit switching guarantees
the best sounding call because all packets go in order. No delay. Delays
in packet switching for voice causes cause voice quality to fall apart,
as anyone who has talked over the internet can tell you. As technology
gets better with time, voice over packet switched networks will get better,
indeed, Bell Labs says that the problems with sending voice over packet
switched networks have been overcome. They don't talk, though, about
sending voice over packet switched networks in a cellular radio context.
Ericsson is confident about the 'air interface' as the following shows:
"Recently, Ericsson and Japan Telecom . . . successfully completed
the world's first field trial of Voice-over-IP [using] wideband CDMA.
The field trial results prove that voice can be efficiently transported
over an IP-based mobile network. This includes the cellular air-interface,
to mobile terminals, with full quality of voice service as well as full
quality of other service features such as data, without loss of capacity.
. . The field trial was conducted in July and August, 2000 with Japan
Telecom at its network center in Chiba, Japan. . . 'The trend in today's
telecoms industry is towards 'all-IP' transport networks," says
Håkan Eriksson, Vice President and General Manger, Ericsson Research.
"Operators want to be able to use the same network for all services;
data, voice and video. The field trial conducted together with Japan
Telecom has proven that it is possible to transport voice over an IP-based
mobile network, without compromising quality or system performance."
Some companies like Caspian
Networks are developing router like devices which will recognize packet
types and prioritize accordingly, thus speeding up packet delivery and
reducing lag time with voice and video. As Josh McHugh writes about Caspian's
optical IP superswitch, in the May 2001 Wired, "It can identify packet
types (voice, text, video, et cetera) and priorities, allowing it to determine
one packet's relation to others, and expedite traffic in a way that's
impossible today. For example, the Aperio will recognize all portions
of a video stream and label them as a part of a greater whole so they
can be more efficiently slotted and moved to their ultimate destination."
We shall see.
Packet switched networks exist for the data communication needs of education,
business, and government throughout the United States. These networks
rely on telephone lines, of course, but the circuits are so arranged that
they retain a permanent connection with their customers. The Public Data
Network or Packet Switched Network, stands as the data counterpart to
the Public Switched Telephone Network. I used to dial a local number to
access Delphi, a now defunct internet service provider. Compu$erve and
Plodigy used the same telephone number. All three used the same packet
network, which you accessed when your computer dialed and logged in. An
identification nmber directed your traffic to the right ISP, no matter
where in the country it was. If you logged out but did not hang up the
modem, you could enter numbers at the prompt on your screen and connect
to, among other services, the NASA packet switching network. But I wander
from the point I wanted to make.
Packet
Video is promising video clips at 60Kbs over conventional circuit
switched cellular radio channels, indeed, they say they are platform
independent, that is, their technology will work over whatever radio
technology a carrier is using. I saw T.V. screen based demo at WirelessIT2000
in Santa Clara, CA recently, although a working device wasn't present.
Unlike circuit switching, no one call takes up an entire channel for
an entire session. Bits get sent only when traffic goes on, when people
actually speak. During pauses in a conversation a channel gets filled
with pieces of other conversations. Because your call doesn't hog an entire
circuit the telephone system can permit an always on connection. You might
pay a flat monthly charge or by the bandwidth or bits you actually use.
Whether wireless operators can afford to do so is difficult to decide.
Too many customers means building many more expensive cell sites. Even
if technology permits we may stay with a per minute charge.
If packet switching is so efficient, why hasn't the landline public
switched telephone network converted to it? The answer is time and money.
Replacing circuit switched switches with packet switches accross the country
would be a monumental task, requiring billions of dollars over years and
years. The legacy of circuit switching will be around for quite a long
time, following us far into the new century. Still, traffic engineers
must think about changing, with lengthy dial up calls to the internet
placing huge demands on switches that were never planned for, circuits
now tied up longer than ever imagined. But change has to come at some
point, and the internet's traffic now motivates engineers to move toward
a unified switching method in the PSTN. As Bell Labs puts it "Telecommunications
companies and Internet providers view these new problems as opportunities
to move from separate voice and data networks to converged packet-switched
voice and data networks."
DSL and ASDL and cable modem connections will either speed or retard
this transistion; a local telephone company directs this broadband traffic
to a packet switch, bypassing the existing local, circuit based switch.
As broadband users increase call holding times should decrease, as dial
up modems are taken out of service. The local switch should not be as
overwhelemed as many currently are. A telco may then decide to delay a
transistion to packet switching.
While the PSTN creeps towards convergence, many telecom companies are
looking at placing calls over packet switched local area networks the
internet. John Quain notes in the October,2000 Computer
Shopper that GTE is partners with Dialpad.com,
a net based service allowing computer to landline telephone calls, while
AT&T owns 30 percent of Net2Phone,
which permits free computer to computer calls. This is voice over internet
protocol technology. Calls sound poor at times, reminding me of short
wave. But free is good, especially if you are an American who needs to
talk with another computer user in New Zealand. Panasonic will soon debut
a cordless phone with a Net2Phone button, push it before making a call
and the cordless will place the connection over the net, with no need
for a computer. Call setup may take a while, of course, but Panasonic
hopes a 3.9 cent a minute toll charge to anywhere in the country will
mollify users. I'm not so sure. Quain also says Netscape's 6.0 browser
has Net2Phone built in but does not say if there is a Macintosh version.
A complete lack of Mac compatible VoIP systems has prevented me from playing
with this technology.
Call quality differs from the PSTN for many reasons: slow speed internet
connections, feedback from poor microphone placement, low grade transmitters
and receivers. Companies using packet switching to place voice calls over
their high speed local and wide area networks don't suffer from these
problems as much. Quain says companies like 3Com market systems to small
firms which funnel inbound calls to the packet switch for a company. Once
packetized the call goes directly to whatever phone number was being dialed.
This eliminates the traditional office switch and allows software, not
hardware, to enable features like conferencing and call forwarding. Even
video conferencing if the number being dialed at the office is to a computer
and not a desk telephone. That's simpler than it sounds.
When a call comes into your computer over such a system a graphic or
an image comes up, saying you have a call. An keypad image lets you point
and click on the numbers to make a call. Your computer or the one for
the company enables voice mail and stores telephone directories. A company
with a packet based switch will alow you to eventually store all of your
e-mail and pages and faxes and voice calls on a single computer which
also acts as your phone. See where convergence is taking us? And how getting
away from circuit switching will help? The drive toward unified packet
switching will enable a brand new future for the public telephone system.
Some people say that Bell System engineers had good ideas for developing
packet switching for voice traffic on the PSTN but I will have to do more
research to confirm this. The following article, written by George Gilder,
gives some clues but no specific references or dates. But for now, knowing
the difference between circuit switching and packet switching will, I
hope, make understanding the new wireless data services a little easier.
|
|