Packet and circuit switching

Have a question or want to start a discussion? Post it! No Registration Necessary.  Now with pictures!

Threaded View
Sometime ago, someone asked about Packet and circuit switching ...
I am just copying and pasting how cisco explains these terms (which are not
very clear to me either :) Notes within the brackets are mine.

Packet-switched networks were developed to overcome the expense of public
circuit-switched networks and to provide a more cost-effective WAN
When a subscriber makes a telephone call, the dialed number is used _to set
switches in the exchanges_ (whatever that might be) along the route of the
call so that there is a continuous circuit from the originating caller to
that of the called party (How does the dialer number do the job of an
operator might be even more complicated). Because of _the switching
operation_ used to establish the circuit, the telephone system is called a
circuit-switched network. (Very nice definition :). If the telephones are
replaced with modems, then the switched circuit is able to carry computer
data. (as clear as mud!)

The internal path taken by the circuit between exchanges is shared by a
number of conversations. Time division multiplexing (TDM) is used to give
each conversation a share of the connection in turn. TDM assures that a
fixed capacity connection is made available to the subscriber. ( isn't this
cool? whatever that might mean :)

If the circuit carries computer data, the usage of this fixed capacity may
not be efficient. For example, if the circuit is used to access the
Internet, there will be a burst of activity on the circuit while a web page
is transferred. This could be followed by no activity while the user reads
the page and then another burst of activity while the next page is
transferred. This variation in usage between none and maximum is typical of
computer network traffic. Because the subscriber has sole use of the fixed
capacity allocation, switched circuits are generally an expensive way of
moving data.

(Is all this stuff in CCNA exam? )

An alternative is to allocate the capacity to the traffic only when it is
needed, and share the available capacity between many users. With a
circuit-switched connection, the data bits put on the circuit are
automatically delivered to the far end because the circuit is already
established. If the circuit is to be shared, there must be some mechanism to
label the bits so that the system knows where to deliver them. It is
difficult to label individual bits, therefore they are gathered into groups
called cells, frames, or packets. The packet passes from exchange to
exchange for delivery through the provider network. Networks that implement
this system are called packet-switched networks.

The links that connect the switches in the provider network belong to an
individual subscriber during data transfer, therefore many subscribers can
share the link. Costs can be significantly lower than a dedicated
circuit-switched connection. Data on packet-switched networks are subject to
unpredictable delays when individual packets wait for other subscriber
packets to be transmitted by a switch.

The switches in a packet-switched network determine, from addressing
information in each packet, which link the packet must be sent on next.
There are two approaches to this link determination, connectionless or
connection-oriented. Connectionless systems, such as the Internet, carry
full addressing information in each packet. Each switch must evaluate the
address to determine where to send the packet. Connection-oriented systems
predetermine the route for a packet, and each packet need only carry an
identifier. In the case of Frame Relay, these are called Data Link Control
Identifiers (DLCI). The switch determines the onward route by looking up the
identifier in tables held in memory. The set of entries in the tables
identifies a particular route or circuit through the system. If this circuit
is only physically in existence while a packet is traveling through it, it
is called a Virtual Circuit (VC).

The table entries that constitute a VC can be established by sending a
connection request through the network. In this case the resulting circuit
is called a Switched Virtual Circuit (SVC). Data that is to travel on SVCs
must wait until the table entries have been set up. Once established, the
SVC may be in operation for hours, days or weeks. Where a circuit is
required to be always available, a Permanent Virtual Circuit (PVC) will be
established. Table entries are loaded by the switches at boot time so the
PVC is always available.

( I do not know what kind of audience CISCO has in mind, but if this stuff
is for a beginner that is trying to study for CCNA, then I am too stupid for
CISCO, or to waste my time with their exams)

The Dude

Re: Packet and circuit switching

Quoted text here. Click to load it

Simplified, but a fair explanation of CIRCUIT vs PACKET switching.......


Consider this network:


Each "node" may have hundreds of ports connected to them,
but only the ports being used to carry information between
the SOURCE and DESTINATION are being shown.

Imagine that within each node the circuits on either side
of the node are physically 'touching' each other, as if they
were patched together via a human operator. (cross-connects)

In such a scenario, any information - data or voice - generated
by the SOURCE would travel down the wire and traverse each cross-connect
within each node until the information ultimately reaches the DESTINATION.

Because because each of the wires are "nailed up" to each other within
the cross-connect ( almost as if they were soldered together ),
there is no need for any type of "address overhead" to accompany
the data/voice information which is traveling down the wires.

Any information generated by the SOURCE simply passes down the physical
path which has been established by each of the cross-connects. None
of this information needs to be inspected in any way by the nodes,
because a physical path has been established between the two end points.
Information entering on one side of each node is simply moved to the other
side of the node.

The cross-connects within each noide can be established in one of two ways:

1 - They are manually established and available 24x7x365. The term for
this is a DEDICATED line. Some people use the term "nailed-up."
leased lines such as a 56k, T1, etc., are examples of this arrangement.

2 - They can be established and dissolved "on demand."
The telephone operator moving patch chords around when making a telephone
call is an example of this.
Of course, we do not employ telephone operators any more to perform this
Instead, we have intelligent "switches" that can be directed via a signaling
mechanism to establish the physical connections for us. The signaling
mechanisms could be rotary dial pulses, DTMF tones from a telephone touch
ISDN Q921/Q931, on hook / off hook conditions, and many others.

Regardless as to if the circuit switched connections were established using
case (1) or case (2), information being passed through each node is never
by the node because there is no need to,
as a 'physical' connection has been estabished and there is no need to
inspect the data content for address / routing information.

What comes in on one side of the node simply goes out the other side of the



There is no physical 'cross-connect' established within each node. Instead,
each parcel of information the SOURCE generates must contain
some_form_of_addressing information. The nodes will inspect this addressing
information and make a forwarding DECISION based upon that addressing
Small 'store and forward' delays are always present within each node. This
because of the time it takes to hold the parcel of information (packet),
the address information found within the packet, and then ultimately forward
the packet along the correct path to the DESTINATION.

There are two generic manners in which to perform packet switching:

1- Datagram service. IP packets fall into this category.
Each packet contains the SOURCE and DESTINATION addresses of the
communication end stations.
Because each packet contains the source and destination addresses,
each packet can be 'independently' routed through each node within the
network - they do not have to follow each other.
An analogy would be what happens after a wedding.
Everyone goes there own different ways, but, somehow some time later,
they all wind up at the same party to eat, drink, and in some cases,
make asses out of themselves. Perhaps too much 'drink' involved.....

2 - Virtual Circuit (VC) mode. VCs operate like a funeral prosession.
The driver of the herse knows where he is going, and all the cars follow the
leader to the final resting place.
VCs tend to preserve the sequential order of packet flow.
Think of a garden hose. If you rolled red, white, and blue marbles into a
garden hose,
then they will show up in the same order - red, white, and blue. (unlike the
'datagram / IP' service where the marbles can possibly show up blue, white,

There are two kinds of Virtual Circuits. "Switched Virtual Circuits / SVC"
and also
"Permanent Virtual Circuits / PVC." In BOTH cases, each packet contains
some_form_of_addressing information to enable to nodes along the path to
decide, and forward the packets along the correct path associated with the
Frame Relay Data Link Connection Identifiers (DLCI), X25 Logical Channel
Numbers (LCN)
are examples of the address information found within each parcell of
information being switched through each node.

In the PVC mode, the path between SOURCE and DESTINATION is pre-determined
the network node administrators, and the DLCI/LCN information is also
pre-determined for that path.
The SOURCE station simply adds the DLCI/PVC information to each parcel being
presented to the network.
The nodes along the path then inspect each parcel, decide where to send
them, and forward them on accordingly.

In the SVC mode (rare with frame relay, common with X25) the path between
SOURCE and DESTINATION is not pre-determined.
Instead there is a form of 'signaling' which is used.
The 'signaling' mechanism provides a way for the SOURCE user to inform the
network that a connection is desired,
and what the desired DESTINATION for the connection is.
Each node takes this 'call request' and establishes a temporary 'on-demand'
VC / garden hose between the SOURCE and DESTANTIONS.
When the signaling mechanism informs the SOURCE user that the
(switched / on-demand) VC has been established, the network will inform the
users what DLCI/LCN to use for the connection.
Packet forwarding will commence with each packet having the appropriate
information appended to them. The nodes will once again inspect this
information, decide what (temporary) connection is associated with these
and then forward the parcells along their way.

When the user no longer needs the on-demand garden hose, the signaling
being utilized informs the network nodes that the connection can be cleared.
The DLCI / LCN assignments which were used for the prior call can then be
back into an available pool of address space for other users to take
advantage of.

The only bad thing about SVC services is that it's a real bithc rolling up
all the garden hoses and putting them back
where they belong after using them. They tend to get tangled up.


Re: Packet and circuit switching
I hear what the Dude is saying-Mr Agosta, your analogys were
excellent-it is so much easier to picture something -many thanks
btw I am undrstanding that the Circuit switched is faster but consumes
mre CPU power? wheras the packet switched is slower but can operate on
a much lower spec machine?
John Agosta wrote:
Quoted text here. Click to load it

Re: Packet and circuit switching

Quoted text here. Click to load it

I wouldn't make those generalizations.
I would say there are usually more 'latencies" involved with
packet switching within each node due to the requirement to accept, inspect,
and then forward the packets.
In a circuit switched environment, things are more like a continuous 'flow,'
resulting in lower latencies.

Re: Packet and circuit switching

Packet switching is making 'direction' decision per individual
data packet. The path over which data  will be flowing is in principle

Circuit switching is making 'direction' decision before any data is
transfered (apart from possible fail-over technique's). The path over
which data will be flowing is the same throughout the whole session (again
apart from possible failover).

And that's it.

There is no concept of "packet switching is faster than circuit switching"
These is a difference between:
    1. setting up the data path.
    2. the bit rate at which data flow over the path.


On Mon, 07 Aug 2006 15:42:14 -0500, John Agosta wrote:

Quoted text here. Click to load it

Re: Packet and circuit switching
Brilliant-the best answer yet!
NO_spamm wrote:
Quoted text here. Click to load it

Re: Packet and circuit switching

Quoted text here. Click to load it

Well, for now (the exam) I have memorized (something I always hated to do)
that Circuit Switched ( which is used in ISDN) is _much slower_ than Packet
Switched ( which is used in Frame Relay). Cell Switched is the fastest and
is used in ATM ....
Obviously, you think the opposite because the explainations are not right. I
do not have much time to look at them today ( as my exam is tomorrow), but I
still appreciate Mr Agosta's reply.

The Dude

Re: Packet and circuit switching

Quoted text here. Click to load it

I think you are confusing the issue of  transport speed vs the switching
ISDN is "slow" because (in BRI) there is 128,000 bps of bearer capability.
That 'speed' is not a limitation of  "circuit switching," but a limitation
of ISDN.
An ISDN connection at 64kb is still faster than say, a 9600 bps packet
switched connection.
ATM can be very "fast" indeed, but this is due to the combination of
hardware-based cell switching,
and the (usually) high speed transport facilities that carry these cells. A
cicuit switched connection at "any"
bit rate offers better throughput/speed than a packet or cell switched
connection over the same medium.

CCNA testing aside, it would be helpful to research/study some of these
basics with respect to telecom/communications.

Re: Packet and circuit switching
Ok- so if you have a higher spec machine that is packet switched it
will go fsater- however the Ciruit swithched should be faster as it is
a direct(more or less) connection?
Clarification would be great-thanks
John Agosta wrote:
Quoted text here. Click to load it

Re: Packet and circuit switching
By faster, are you referring that it uses less CPU/memory or that it is
capable of transferring more data per second?  If you are going by the
CPU/memory idea, then circuit switching is a little bit simpler, and
consumes less resources.  As for the transferring of data, packet
switching can be more efficient because it can use multiple links, make
better decisions on which links to use, etc etc.  But this comes at a
cost, which is higher processor usage, more memory, and the added
complexity of making an IP network.  Like for instance if you're using
OSPF as a routing protocol, it can use a lot of resources.  It also
doesn't have to establish and set up a dedicated circuit to the
destination, which causes some overhead.

I'm still studying for the CCNA exam, so take my words with a grain of salt.

Re: Packet and circuit switching

John Agosta wrote:
Quoted text here. Click to load it

John Agosta wrote:
Quoted text here. Click to load it

thanks, but what does the word "switching"  mean in telecommunications
(no doubt it's an old term - I suppose predating cisco's naming their
products "switches"?

Passing the "packet"(as in the whole lot) from one telecommunications
device to another?  I doubt it's called switching and at layer 2, and
for that reason.  If that were so, surely they'd use the old word for
that function - bridging. And if it were at layer 3 it'd be called

Re: Packet and circuit switching

Quoted text here. Click to load it

I don't quite understand your question/statement in the first two sentences
of the last paragraph,
but I will attempt to discuss.....

In the old telcomm world, the word "switching" usually involved layer 1
IE - mapping bits from one channel into another. For simplicity's sake, we
could call this
a "cross connect," or a "DCS." ( digital cross connect system). Sometimes
called "DACS."
(A for "access")

Assuming the DCA is not in some alarm state,
your typical DCS would not "inspect" the contents of a bit stream to
determine any intelligent
(address/protocol/content) information. Bits would simply come in one side
of the DCS and exit out
the other side. The mapping of input-to-output channles would be done on an
administrative staic
basis, or, with some signaling mechanism - which in itself is another topic.

Enter the world of 'computer communications,' and for instance, ethernet
driven network devices.

Here, the word "switching" includes some "inspection" of the content /
 Layer 2 inspection - like MAC addresses.

The term "routing" would infer the inspection of the 'layer 3' contents
carried within the layer 2 'frame.'

Suggest you read slowly the elaboration about  "circuit vs packet" switching
in the previous post
because it is pretty well spot-on.

Re: Packet and circuit switching

John Agosta wrote:
Quoted text here. Click to load it

it looks like you did !

Quoted text here. Click to load it

great, as I expected

Quoted text here. Click to load it

thanks,    I googled around but didn't find anything. Is there basic
resource on this?

it's obviously a network of nodes and the packet gets switched between

talking telecommunications-
So, The whole system is called DCS or DACS
What is the name of the nodes(i.e. it's not a router or bridge, it's a
what?), and is it correct to say that the packet is "switched between
i.e. "..the packet is being switched between the ____ (the what?!) "

i'm sure cisco doesn't go into this at that level. But at the level of
what is the whole system, and what are the nodes that the packet passes
through, I think that's a minimal level for a thinking individual!

Quoted text here. Click to load it

yeah, i know that one a bit, lots of info available for that one!


Re: Packet and circuit switching
Perhaps the following would be a good starting point to bone up on some the
circuit switched (T1) physical layer issues......

Quoted text here. Click to load it

Re: Packet and circuit switching

John Agosta wrote:
Quoted text here. Click to load it
Quoted text here. Click to load it

thanks, I will study that

Re: Packet and circuit switching
On Wed, 16 Aug 2006 12:53:33 -0700, wrote:

Quoted text here. Click to load it

I think you should let go the mapping between the OSI layers and the
naming. Why? How would you handle/name the following technologies?

Label switching, lambda switching, L3 switching, L4 switching, cable

Every technology should be considered in its context. You can have TCP
over IP over GRE inside IPsec over a MPLS infrastructure
over IP over Frame Relay over SDH over lambda (colors) over fibre. So,
how are you naming each of the encapsulations?

And if I add RSVP for bandwidth allocation inside the MPLS infrastructure,
I am bringing in a 'kind of circuit switching technology' into the story.

If you look at history out technologies started out to be circuit switched
(Bell started with this :-)), we moved to packet switching, but we like to
have bandwidth guarantees (for voice, for example), so, we build
technologies (algorithms like Marking, WRED, CBWFQ, LLQ, etc) to
simulate circuits switch technologies on top of a packet switched


Re: Packet and circuit switching
Quoted text here. Click to load it
No_spamm wrote:
Quoted text here. Click to load it

I wasn't really trying to put it in an OSI layer.  I was really
determining that in this context, switching wasn't xyz.

Quoted text here. Click to load it

That is a great demonstration of where forcing the OSI layer
terminology can be inappropriate. If I want to look into all those
technologies, to get a good overview, what is a good route  - any
particular resource(s) or cert track?

So far i'm only really aware of the CCNA for a broad view of
networking.  CCNP then going more specific in it. What would cover the
kind of things you mention?

many thanks

Re: Packet and circuit switching
On Sat, 19 Aug 2006 16:19:58 -0700, wrote:

Quoted text here. Click to load it

I am not really familiar with the latest certifications contents.
But I do remember that CCNA talks about some general architectures, like
the OSI, model. On CCNP, I am not sure (skipped that one...). I believe
CCNP gives you some more details on how several routing protocols work.
However, I once met a CCNP certified instructor. And he was convinced
that in order to have BGP in your network, you do not need any underlying
IGP (OSPF, EIGRP, RIPv2, ISIS, whatever). That's a bad instructor...

Anyway, I suggest doing the following:

Try to understand the OSI model, just for the context. I mean the
relationship between the layers. Then try to understand how for example
a 'common' TCP/IP stack is built and where it (relatively) fits into the
OSI model. There are probably some drawings out there that show the
relationship of the several protocols, just an example:

UTP, 802.3, IP, TCP, telnet
fiber, 802.3, IP, TCP, telnet
fiber, SDH/SONET, IP, TCP, telnet
T1, Frame-relay, IP, TCP, telnet
T1, HDLC, IP, TCP, telnet

Then try to understand some individual protocols (fitting your CCNA
cert.) and fit them into the picture. When you look at the header of these
protocols, you'll see some "fields" that refer to the next encapsulation
layer. Don't let the architecture itself scare you of: MPLS labels are
sitting between the MAC header and IP header (or layer 2 and layer 3). But
how labels are actually distributed (together with label binding, label
stacking, etc) is another story and may depend on the underlying
technology (Cell-mode MPLS on ATM versus Frame-mode MPLS on non ATM, but
you can have frame-mode MPLS over ATM :-). BTW: the difference is whether
you use VPI/VCI cell identifiers in ATM instead of Labels and use a
"VPI/VCI - Label mapping" at the head-ends). Anyway, that's why MPLS is
sometimes called a layer 2.5 protocol.


Re: Packet and circuit switching

NO_spamm wrote:
Quoted text here. Click to load it

thanks, I know the OSI and have seen fields that refer to the protocl
above e.g. Ethernet has Type , which can refer to IP or IPX etc.  But
CCNA doesn't offer an overview of so many technologies.

I think CCNA is probably a bit weak on the technology section.In that
it's not broad or deep. Especially AWN technologies.

CCNP i've heard isn't much broader than CCNA. Just more depth. So it
won't have the breadth either. Which might explain that bad instructor.

OSI isn't a problem.

Any ideas on how to see/run into/learn about those technologies though
- getting at least a good overview?

Re: Packet and circuit switching
On Sun, 20 Aug 2006 14:35:27 -0700, wrote:

Quoted text here. Click to load it

Sounds silly but:
1. Google
2. Draw pictures yourself. That's what I did.
3. Read a lot:
    Cisco website (not always clear, btw),
    RFCs (yes, sometimes very very useful if you want to know
        exactly how it works)
    Cisco Press books (sometimes many errors)


Site Timeline