question about networking hubs / ethernet repeaters

[[email protected]]

question about networking hubs / ethernet repeaters -

say you have comps A,B,C,D connected to a hub.

It's my understanding that the signal from a comp will not be sent back
to that comp.

A hub uses a bus/ just one wire to get the hub effect of sending to
all. What electronic wizardry prevents a comp from receiving its own
signal or more correctly, prevents the hub from sending the signal back
to the orgiinating comp?

 

[Tom Biasi]

question about networking hubs / ethernet repeaters -

say you have comps A,B,C,D connected to a hub.

It's my understanding that the signal from a comp will not be sent back
to that comp.

A hub uses a bus/ just one wire to get the hub effect of sending to
all. What electronic wizardry prevents a comp from receiving its own
signal or more correctly, prevents the hub from sending the signal back
to the orgiinating comp?

The hub will send the packets to all devices connected including the
originating device.

 

[[email protected]]

The hub will send the packets to all devices connected including the
originating device.

can you give a reference for that?

I will give a reference for what I said, which is that the hub sends to
all except the original device

p.60 "The hub's internal wiring propagates the signal to all other
ports, but not back to the port that the signal was received on"
CCNA Intro by Wendell Odon CCIE No 1624

 

[Andrew Holme]

question about networking hubs / ethernet repeaters -

say you have comps A,B,C,D connected to a hub.

It's my understanding that the signal from a comp will not be sent back
to that comp.

A hub uses a bus/ just one wire to get the hub effect of sending to
all. What electronic wizardry prevents a comp from receiving its own
signal or more correctly, prevents the hub from sending the signal back
to the orgiinating comp?

A dumb repeater hub connects stations together to form a half-duplex
collision domain in which only one station can transmit at a time. A
collision occurs if two stations transmit at once. Stations need to be able
to monitor their own transmissions in order to detect collisions.

An intelligent switching hub routes packets only to the intended recipient
by inspecting the destination MAC address field in the packet header. Full
duplex operation is supported. Stations do not receive their own
transmissions.

 

[[email protected]]

A dumb repeater hub connects stations together to form a half-duplex
collision domain in which only one station can transmit at a time. A
collision occurs if two stations transmit at once. Stations need to be able
to monitor their own transmissions in order to detect collisions.

"Stations"/devices monitor their own transmissions without the hub
sending their own transmission back to them. Statons use a loopback
circuit in their own NIC (a connection between transmit and receive
wires) thus, sending their own signal back to them - but not via the
hub. Collisions are detected of course.
When there is a collision, the transmitting stations will detect it.
Station A which is receiving its own signal via loopback, will receive
B's signal via the hub. . The combination causes a detectable change in
the electrical signal (like increase voltage or maybe amplitude) they
they receive a signal, from the hub, from that other computer. The
same that goes for Station A goes for Station B.

An intelligent switching hub routes packets only to the intended recipient
by inspecting the destination MAC address field in the packet header. Full
duplex operation is supported. Stations do not receive their own
transmissions.

i'm not talking about switches. A switch is a hub, as much as a router
is a hub. It's in the same sense, as in "hub and spoke topology". But
I am not talking about hub in the sense of "hub and spoke topology".
Nor am I talking about further marketting with terms like "intelligent
hub". I only mean specifically (multiport) "ethernet" repeaters.

 

[John]

"Stations"/devices monitor their own transmissions without the hub
sending their own transmission back to them. Statons use a loopback
circuit in their own NIC (a connection between transmit and receive
wires) thus, sending their own signal back to them - but not via the
hub. Collisions are detected of course.
When there is a collision, the transmitting stations will detect it.
Station A which is receiving its own signal via loopback, will receive
B's signal via the hub. . The combination causes a detectable change in
the electrical signal (like increase voltage or maybe amplitude) they
they receive a signal, from the hub, from that other computer. The
same that goes for Station A goes for Station B.


i'm not talking about switches. A switch is a hub, as much as a router
is a hub. It's in the same sense, as in "hub and spoke topology". But
I am not talking about hub in the sense of "hub and spoke topology".
Nor am I talking about further marketting with terms like "intelligent
hub". I only mean specifically (multiport) "ethernet" repeaters.

In your terminology, any device with two or more connections could be
called an "Ethernet repeater". However, a switch is differentiated
from a hub by the way it handles packets. Switches, like routers,
have memory to be able to send packets only to the desired destination
port. Hubs, by definition, are party line devices.

Until you see that difference, none of your questions are meaningful.

No, I don't have a CCNA - my job was fixing the problems the CCNA
holders created...

 

[jasen]

question about networking hubs / ethernet repeaters -

say you have comps A,B,C,D connected to a hub.

It's my understanding that the signal from a comp will not be sent back
to that comp.

a hub will send it back a switch will figure out where it's supposed to go
and only send it there.

A hub uses a bus/ just one wire to get the hub effect of sending to
all.

ethernet was originally implemented with coaxial cables and used two
conductors to carry the signals (the inner and the shield of the coax)
now with twisted pair cables a separate pair is used for carrying traffic in
each direction.

pergaps you mean one cable?

What electronic wizardry prevents a comp from receiving its own
signal or more correctly, prevents the hub from sending the signal back
to the orgiinating comp?

from a hub nothing prevents that. the originating computer is configured to
ignore its own signals, that's all.

with a swtich the switch "asks around" and then remembers where the packet
is supposed to go to.

these days most hubs are switches.

Bye.
Jasen

 

[Tom Biasi]

can you give a reference for that?

I will give a reference for what I said, which is that the hub sends to
all except the original device

p.60 "The hub's internal wiring propagates the signal to all other
ports, but not back to the port that the signal was received on"
CCNA Intro by Wendell Odon CCIE No 1624

I'm a little uncertain about your level of understanding but you can start
here: http://compnetworking.about.com/od/hardwarenetworkgear/l/aa012801a.htm
The text you quote is an excellent book on networking but assumes you can
get the nuances.
Regards,
Tom (CCIE)

 

[[email protected]]

In your terminology, any device with two or more connections could be
called an "Ethernet repeater".

whose terminology? I never said such nonsense. On the contrary. I was
clear that when I say hub here, I was referring to "ethernet"
repeaters. I was not referring to the "hub and spoke topology"
meaning. I even specified what I meant in the subject. It was another
poster that brought up about "intelligent hubs" (he meant switches).

However, a switch is differentiated
from a hub by the way it handles packets.

switches aren't aware of layer 3, certainly hubs aren't. But yes, of
course they handle it differently. I am well aware that a hub transmits
to all (except the original) , and a switch only to the specified
destination(s). switches read the MAC address. Hubs aren't aware of
that. I never said anything to contradict that, I suggest you read
carefully.

Switches, like routers,
have memory to be able to send packets only to the desired destination
port. Hubs, by definition, are party line devices.

"party line device" must be an old term because it doesn't get many
results on google. I suppose you mean hubs send to everybody (except
the origianal?) .
I'm well aware of that, refer to the question.

Until you see that difference, none of your questions are meaningful.

I was absolutely clear with what I meant by hub. The way I used it ,
It can ONLY mean ethernet repeater. Notice that in my subject, I
referred to ethernet repeater..

I only mentioned the concept of the term hub in "hub and spoke
topology" when a poster mentioned the term "intelligent hub"(he
probably meant 'switch). I brought that up to indicate that I did not
mean switches. Or routers for that matter

Now, i'm hoping somebody can answer by what electronic wizardry, hubs
("ethernet" repeaters) do not send back to the originating device.
That is a meaningful question
Or do you disagree with CCIE Wendell Odom and say that they do send
back to the originating device?


If they do send back to the originating device , then what happens with
a hub connected to a hub, wouldn't it go into a loop?

 

[[email protected]]

I'm a little uncertain about your level of understanding but you can start
here: http://compnetworking.about.com/od/hardwarenetworkgear/l/aa012801a.htm
The text you quote is an excellent book on networking but assumes you can
get the nuances.
Regards,
Tom (CCIE)

I understand the nuances(e.g. in Odom's discussion of hubs), and
certainly am well aware of what that article discusses.

I quoted from CCIE Wendell Odom's book, you say it's an excellent book.
Yet It flatly disagrees with you. I don't see how that article you
refer to is relevant


You-
The hub will send the packets to all devices connected including the
originating device.

Odom-
"
p.60 "The hub's internal wiring propagates the signal to all other
ports, but not back to the port that the signal was received on"
CCNA Intro by Wendell Odon CCIE No 1624
"

Also, if you are correct (and odom wrong), then what happens when a hub
sends to a hub, would it not loop?
By Odom's explanation, it wouldn't, so is not a problem. By yours, it
would be. But you seem to not yet acknowledge that there is even a
disagreement between you and Odom here.

 

[[email protected]]

a hub will send it back a switch will figure out where it's supposed to go
and only send it there.


ethernet was originally implemented with coaxial cables and used two
conductors to carry the signals (the inner and the shield of the coax)
now with twisted pair cables a separate pair is used for carrying traffic in
each direction.

pergaps you mean one cable?

I said, by hub, I mean ethernet repeater.THat is not a switch. Data
will not flow in 2 directions. "Ethernet" repeaters are half duplex,
not full duplex.

from a hub nothing prevents that. the originating computer is configured to
ignore its own signals, that's all.

ok, so you disagree with Odom then.

so, as i've asked others, what of when 2 hubs are connected, how is a
loop prevented? (i've asked the q of anybody that said this, but they
haven't posted yet prob 'cos it's nighttime!)

with a swtich the switch "asks around" and then remembers where the packet
is supposed to go to.

I didn't mean switches. only ethernet repeaters.

these days most hubs are switches.

I didn't mean them.

 

[Tom Biasi]

I understand the nuances(e.g. in Odom's discussion of hubs), and
certainly am well aware of what that article discusses.

I quoted from CCIE Wendell Odom's book, you say it's an excellent book.
Yet It flatly disagrees with you. I don't see how that article you
refer to is relevant


You-
The hub will send the packets to all devices connected including the
originating device.

Odom-
"
p.60 "The hub's internal wiring propagates the signal to all other
ports, but not back to the port that the signal was received on"
CCNA Intro by Wendell Odon CCIE No 1624
"

Also, if you are correct (and odom wrong), then what happens when a hub
sends to a hub, would it not loop?
By Odom's explanation, it wouldn't, so is not a problem. By yours, it
would be. But you seem to not yet acknowledge that there is even a
disagreement between you and Odom here.

I don't have a copy of that book handy to see the context in which that
statement was made.
You seem to realize that the hub doesn't read any data at all. If it doesn't
read data it can't make any decisions. As someone pointed out, the computer
is configured to ignore its own signal (echo-off).
The hub doesn't store any data before transmitting so a loop is meaningless.
BTW: For someone looking for help you could be a little more polite.
Tom

 

[[email protected]]

I don't have a copy of that book handy to see the context in which that
statement was made.

I understood him correctly.

he states the same in another book too. in computer networking first
step.
"the hub simply listens for incoming electrical signals and whne
received the hub repeats the same electrical signal to every other
device that's conected to the hub........."
"the hub's logic is simple.... receive traffic on pins 1 and
2.....repeat the signal to all other ports except the one in which the
data was received..when erpeating out other ports, repeat out traffic
on pins 3 and 6"

And his concept of collision detection relies on that..
Here is another quote from CCNA Intro which makes Odom's concept of
what happens very clear. p.60
"If PC1 and PC2 sent a frame at the same time , a collision would
occur. .. The hub would forward both electrical signals, which would
cause the overlapping signals to be sent to all the NICs. So, because
collisions can occur, the CSMA/CD logic still is needed to have PC1 and
PC2 waited and try again. Note - PC2 would sense a collision because of
its loopback circuitry on the NIC. THe hub does not forward the signal
that Pc2 sent to hte hub back to PC2. INstead, each NIC loops the frame
that it sends back to tits own receive pair on the NIC.. Then if PC2 is
sending a frame and PC1 also sends a frame at the same time, the signal
sent by PC1 is forwarded by the hub t oPC2 on PC2's receive pair. THe
incoming signaal from the hub, plus the looped signal on PC2's NIC,
lets PC2 notice that there is a collision. "

You seem to realize that the hub doesn't read any data at all. If it doesn't
read data it can't make any decisions. As someone pointed out, the computer
is configured to ignore its own signal (echo-off).

hmm, I guess even regularly, when it uses loopback, it ignores its own
signal, even though the signal comes in on the receive pairs. But
how?.

I asked in an electronics group hoping to find out by what electronic
wizardry. Or does the NIC ignore based on the MAC?
I guess that'd be how it happened with the old coax LAN technologies ,
that didn't use loopback.

The concept people discuss here is clearly different to that discussed
by Odom. Fair enough.

What is the purpose of loopback then?

The hub doesn't store any data before transmitting so a loop is meaningless.

even without storing data. The wire between the hubs would be
permanently busy.

BTW: For someone looking for help you could be a little more polite.
Tom

Sorry if it appears that way, but I am writing in this manner in order
to get facts straight. Were I to not write in this manner, emphasising
where the disagreements are, then things wouldn't be so clear. Infact,
it may have gone unnoticed that every poster here disagrees with Odom.

I hope that the extra quotes from Odom make Odom's position / the
disagreement clearer. I'd be interested if you can recall where you
get your fact that the hub sends to all devices and the originating
comp ignores it. Because I have a source for my fact that the hub
sends to all devices except the originating device.

thanks

 

[Andrew Holme]

I said, by hub, I mean ethernet repeater.THat is not a switch. Data
will not flow in 2 directions. "Ethernet" repeaters are half duplex,
not full duplex.



ok, so you disagree with Odom then.

so, as i've asked others, what of when 2 hubs are connected, how is a
loop prevented? (i've asked the q of anybody that said this, but they
haven't posted yet prob 'cos it's nighttime!)


I didn't mean switches. only ethernet repeaters.


I didn't mean them.

If I wanted to make a crude 4-port repeater-type hub for 10BASE-T, using
only simple logic gates, with no microprocessor, no FPGA and no signal
re-timing even, the logic might go something like this:

Output[0] = Input[1] OR Input[2] OR Input[3]
Output[1] = Input[0] OR Input[2] OR Input[3]
Output[2] = Input[0] OR Input[1] OR Input[3]
Output[3] = Input[0] OR Input[1] OR Input[2]

Each output is generated from the logical OR of all other inputs except it's
own.

 

[Tom Biasi]

I understood him correctly.

he states the same in another book too. in computer networking first
step.
"the hub simply listens for incoming electrical signals and whne
received the hub repeats the same electrical signal to every other
device that's conected to the hub........."
"the hub's logic is simple.... receive traffic on pins 1 and
2.....repeat the signal to all other ports except the one in which the
data was received..when erpeating out other ports, repeat out traffic
on pins 3 and 6"

And his concept of collision detection relies on that..
Here is another quote from CCNA Intro which makes Odom's concept of
what happens very clear. p.60
"If PC1 and PC2 sent a frame at the same time , a collision would
occur. .. The hub would forward both electrical signals, which would
cause the overlapping signals to be sent to all the NICs. So, because
collisions can occur, the CSMA/CD logic still is needed to have PC1 and
PC2 waited and try again. Note - PC2 would sense a collision because of
its loopback circuitry on the NIC. THe hub does not forward the signal
that Pc2 sent to hte hub back to PC2. INstead, each NIC loops the frame
that it sends back to tits own receive pair on the NIC.. Then if PC2 is
sending a frame and PC1 also sends a frame at the same time, the signal
sent by PC1 is forwarded by the hub t oPC2 on PC2's receive pair. THe
incoming signaal from the hub, plus the looped signal on PC2's NIC,
lets PC2 notice that there is a collision. "




hmm, I guess even regularly, when it uses loopback, it ignores its own
signal, even though the signal comes in on the receive pairs. But
how?.

I asked in an electronics group hoping to find out by what electronic
wizardry. Or does the NIC ignore based on the MAC?
I guess that'd be how it happened with the old coax LAN technologies ,
that didn't use loopback.

The concept people discuss here is clearly different to that discussed
by Odom. Fair enough.

What is the purpose of loopback then?



even without storing data. The wire between the hubs would be
permanently busy.



Sorry if it appears that way, but I am writing in this manner in order
to get facts straight. Were I to not write in this manner, emphasising
where the disagreements are, then things wouldn't be so clear. Infact,
it may have gone unnoticed that every poster here disagrees with Odom.

I hope that the extra quotes from Odom make Odom's position / the
disagreement clearer. I'd be interested if you can recall where you
get your fact that the hub sends to all devices and the originating
comp ignores it. Because I have a source for my fact that the hub
sends to all devices except the originating device.

thanks

OK, I thought that what I sensed as impoliteness may have been frustration.
I have seen Odom state what you say many times and I have seen hub
manufacturers circuits that retime, amplify and provide one big collision
domain. The nic is responsible for managing collisions. As you surmise the
wiring is quite busy. Collisions are happening all the time on a hubed
network. CSMA/CD protocol manages this situation. The definition of a
simple dumb hub is changing every day. Hubs are taking on some of the
features of switches and some have a side processor to do what you said. A
true dumb hub doesn't know which port to exclude. In half duplex its fairly
easy; it cant talk and listen at the same time, in full duplex it needs the
nic and CSMA/CD to handle collisions.

Let me conclude by saying, I really can't answer your question as it was
stated.
Sorry,
Tom

 

[[email protected]]

OK, I thought that what I sensed as impoliteness may have been frustration.
I have seen Odom state what you say many times and I have seen hub
manufacturers circuits that retime, amplify and provide one big collision
domain. The nic is responsible for managing collisions. As you surmise the
wiring is quite busy. Collisions are happening all the time on a hubed
network. CSMA/CD protocol manages this situation. The definition of a
simple dumb hub is changing every day. Hubs are taking on some of the
features of switches and some have a side processor to do what you said. A
true dumb hub doesn't know which port to exclude. In half duplex its fairly
easy; it cant talk and listen at the same time, in full duplex it needs the
nic and CSMA/CD to handle collisions.

Let me conclude by saying, I really can't answer your question as it was
stated.
Sorry,
Tom

thanks for trying. no need to be sorry. (As you should know. Since it
is obvious that you did what you can and you didn't even need to do
that)
And I wasn't frustrated. It was merely as I said, I was making the
disagreement clear so as to clarify the question. I will try in
sci.electronics.repair , hopefully some people there have opened
hubs up and perhaps even seen both , ethernet repeaters that do send
to all, and ethernet repeaters that send to all except the originating.
I do have a suspicion that all of them don't send to the originating.

 

[[email protected]]

I said, by hub, I mean ethernet repeater.THat is not a switch. Data
will not flow in 2 directions. "Ethernet" repeaters are half duplex,
not full duplex.



ok, so you disagree with Odom then.

so, as i've asked others, what of when 2 hubs are connected, how is a
loop prevented? (i've asked the q of anybody that said this, but they
haven't posted yet prob 'cos it's nighttime!)


I didn't mean switches. only ethernet repeaters.


I didn't mean them.

If I wanted to make a crude 4-port repeater-type hub for 10BASE-T, using
only simple logic gates, with no microprocessor, no FPGA and no signal
re-timing even, the logic might go something like this:

Output[0] = Input[1] OR Input[2] OR Input[3]
Output[1] = Input[0] OR Input[2] OR Input[3]
Output[2] = Input[0] OR Input[1] OR Input[3]
Output[3] = Input[0] OR Input[1] OR Input[2]

Each output is generated from the logical OR of all other inputs except it's
own.

so are you saying that there's one chip iin the hub that "closes off"
all ports , only "opening" and outputting to them when another port is
sending.

the bus in the hub is a single wire. Where is the chip located to have
that effect? or is it chips?

(I suppose the single wire in the hub can double up to connect to the
"twisted pair" cables plugged into it. But that's just the end of each
port on the hub, not really relevant. Or it could be that the hub has
a TP wire all the way through it, but again, that doesn't change this.
Though would be interesting to know)

thanks in advance