Collisions domain

[gattina59]

Hello everyone on a question ccna I found different answers and I noticed there is too much confusion seems that the answer is subjective ……

If I have two switches connected in the same broadcast domain then switch A link switch B we have 3 hosts connected.

There is obsiously a link between switch A and Switch B.

The answers are different there are those who say 7 and there are those who say 6. Who says 6 also gives the motivation; He says why the switch to switch connection is full duplex and when it is full duplex is not considered.

The following statement is written in the ccna book: In a modern Lan with all switches and routers, even thought full duplex removes collisions, think of each link to a separate collision domain when the need to troubleshoot arises.

Who should we believe?

[Giuseppe Larosa]

Hello,

each port of a modern transparent bridge ie. switch is a different collision domain regardless of duplex settings.

So if you have two LAN switches with 3 hosts each and a link between them the total number of collision domains is seven and not six.

The concept of collision domain as the concept of broadcast domain has to be intended in the following way: who is involved in a possible collision if the port would work in half duplex mode both ends?

Because a modern LAN switch has a switching fabric inside it is able to switch many frames at the same time between different pairs of ports. This is the technical reason for saying that each switch port is a separate collision domain: there is no chance of having collisions inside the switch itself as it works in hardware.

Hope to help

Giuseppe

[gattina59]

Sembra tu sia Italiano ecco forse megli oscriverlo in Italiano forse mi sono spiegata male io :D.

Guarda questo link https://www.certificationkits.com/cisco-certification/ccent-640-822-icnd1-exam-study-guide/cisco-ccent-icnd1-640-822-exam-certification-guide/cisco-ccent-icnd1-internetworking-and-security-part-ii/  qui ti dice alla seconda immagine che sono 6 non 7 quindi switch a switch non lo considera in quanto full duplex.

Guarda ora questo link  http://www.9tut.com/ccna-basic-questions-2/comment-page-17   domanda 8 stesso scenario me ne considera 7 e non 6 e queste tieni presente sono domande di esame.

Cioè perchè in rete c'è questa confusione? Io all'esame cosa considero?

[Giuseppe Larosa · Accepted solution]

Ciao,

this is an english based site so we should write in english.

>> Cioè perchè in rete c'è questa confusione? Io all'esame cosa considero?

Innanzitutto, entrambi i link a cui fai riferimento non sono siti ufficiali Cisco dovresti considerare la Cisco Learning Network per questo.

Detto questo, ogni porta di uno switch è un collision domain quindi la risposta corretta per me è 7, perchè il link tra i due switch lo conto una volta sola non due volte.

Inoltre, la capacità di creare un collision domain per porta dello switch deriva da come funziona lo switch internamente e non dal duplex del link. Questo è concettualmente sbagliato.

Mi sembra sbagliato contarlo zero volte come accade nel primo link.

### english version ####

Both links you have provided are not official Cisco material, you should consider the cisco learning

network for study.

In my humble opinion the number of collision domains is 7 because the switch to switch link may be working in full duplex, but the capability to provide collision domain separation is related to what is INSIDE the switch, not the duplex state of the link, and the link has to be counted only one not two.

Hope to help

Giuseppe

[gattina59]

Si ok è chiaro però immagino tu abbia fatto un esame ccna non ti è capitata con due switch?; ora bisogna capire come verrà posta la cosa sull'esame.

Queste due frasi prese dal libro cisco che ho:

1) A modern LAN, with all LAN switches and routers, with full duplex on each link, would
not have collisions at all.  
Nella prima frase infatti usa il condizionale;

E sulla base di questa subito dopo dice:

In a modern LAN with all switches and routers, even though full duplex removes collisions,
think of each Ethernet link as a separate collision domain when the need to troubleshoot
arises.

Quindi è come dici tui router e gli switch ragionano che ogni porta è una collisione.

Però tra due collegamenti wan non c'è collisione.

Ora devo solo capire se nell'esame viene conteggiato. Il ragionamento che tu fai è corrispondente a quanto scritto ma capisci bene che devo fare un'esame non voglio essere bocciata :)

[Giuseppe Larosa]

Ciao,

sì ho fatto l'esame CCNA nel 2000 però :).

Credo che a livello di esame CCNA siano importanti i concetti e quindi il fatto che ogni porta di switch sia un collision domain. Questo perchè non può esserci collisione tra frame ricevute su porte differenti di uno switch per le capacità di commutazione simultanea, ed il fatto che lo switch propaga le frame in modo intelligente perchè apprende i MAC address.

Al contrario un hub, è un ripetitore elettrico a livello fisico se un apparato trasmette il segnale è replicato elettricamente su tutte le porte e tutti gli apparati connessi ad un hub operano in half duplex e devono contendersi la banda del collision domain.

Quando una NIC è in half duplex confronta il segnale che invia con quello che riceve dall'hub, se ci sono differenze comprende che ci è  stata collisione cioè sovrapposizione di segnali tra due stazioni.

Vedi su wikipedia

https://en.wikipedia.org/wiki/Network_switch

# english version #############

Thanks for you kind remarks

For CCNA level exam what is important are the concepts and the basis of networking: a switch port provides a separate collision domain, a router port provides a separate broadcast domain AND a separate collision domain.

Clearly in a point to point link operating in full duplex mode collisions do not happen.

However, the speed of a link is negotiated  between the two devices, the duplex should also be negotiated but not all the times this happen successfully.

see

https://en.wikipedia.org/wiki/Autonegotiation

If the speeds are diffferent the link does not come up, but the link comes up even if there is a duplex mismatch or the two NICs are not able to negotiate the duplex settings.

(see interoperability issues section in the link above)

But in my opinion this is more a question for the CCNP TSHOOT exam, as the interoperability issues between Cisco devices and other vendors and between Cisco devices of different ages is a more advanced topic.

Hope to help

Giuseppe

[Joseph W. Doherty]

"In my humble opinion the number of collision domains is 7 because the switch to switch link may be working in full duplex, but the capability to provide collision domain separation is related to what is INSIDE the switch, not the duplex state of the link, and the link has to be counted only one not two."

Although marked as an accepted solution, I still propose that the maximum number of collision domains, assuming each of the 3 hosts has a direct port connection, and the two switches have a interconnection, is four.

Again, collision domains arise in "shared wire" Ethernet.  Any NIC on such a medium has the chance to transmit at the same time as another NIC which then creates a collision.

Replace the two switches with two hubs and how many collision domains are there? A. just one.

Replace the two hubs with two dumb switches, with all interfaces still operating in half duplex mode (as they did with the hubs), and you have four collision domains. The "shared wires" are between each host and each switch port, and between the two switch ports.

Imagine two of those hosts transmitting data, concurrently, between themselves. If on the same switch, a host and the switch it connects to, each could transmit on that shared wired at the same time allowing for a collision.  Again, this is if using half duplex mode.  Two hosts, same switch, creates two collision domains. If hosts are on different switches, you've created another "shared wire", the link between the switches.

I don't recall when the first switches came on the market, whether they all offered any full duplex ports. Even if they did, host NICs at the time did not. Yet the switch was still much better than a hub.

Remember a 48 port hub had one collision domain, and all unicast traffic was seen on all ports.

On a 48 port switch, to the same half duplex capable hosts, offered 48 separate collision domains, and unicast traffic did not flood to all the ports.

Before switches, you could improve network performance by using L2 bridges to segment shared Ethernet segments. Each bridge would separate collision domains.

The first switches were also known as multi-port bridges. They "microsegmented" "shared wire" segments, i.e. each host could have its own collision domain.

BTW, when switches offered full duplex, it was often only used between switches, again, because at the time, most host NICs did not support it. However, as an inter switch link might be the busiest link of the switch, having the link no longer being a collision domain, i.e. the two switches could transmit and receive frames without collisions, was often a net improvement.

If it's difficult to understand a collision domain on a hub (or switch using half duplex), you might want to review even earlier Ethernet using 10Base5, or 10Base2, where the wire between hosts was also physically shared.

For a fun thought, consider what would be better between switches or switch to host, 10 full duplex (no collisions) or 100 half duplex (possible collisions).

[Joseph W. Doherty]

Oh, something else I recall doing, back when switches first came on the market, which may further clarify collision domains. Assume you have one switch to which you connect four hubs. How many collision domains does this make? Does number of hosts, per hub, impact the number of collision domains?

[Teacher Todd Booth]

I'm so sorry, it was a typo, but based on the typo, I agree with you.  I'll try to correct the typo.

[Joseph W. Doherty]

To muddy the waters further, if all the links are full duplex, there are no collision domains. If all the NICs are half duplex, there are four collision domains. (The forgoing assumes each host has its own direct connection to a switch port, and the two switches are interconnected with a non-shared direct link too.)

A collision domain arises for each "wire" (not host NIC) being shared by multiple hosts in half duplex mode.

If 3 hosts were all on the same "shared wire" (like a hub), there would be but one collision domain.

If you place 2 hosts on one shared wire, connect that to a bridge which connects to another shared wire with 3 hosts, you would have two collision domains (and one broadcast domain).

3 hosts, each on its own wire to a multi port bridge (e.g. dumb switch), would be 3 collision domains (and still one broadcast domain).

Move one of those hosts to another (dumb) switch adds only a 4th shared wire.

[Teacher Todd Booth]

To understand why so many posts are incorrect, simply consider the following:

Q1: If there are two nodes, connected directly via half duplex, with the same wires, to a switch, and the switch is powered on, how many collision domains are there?  A: 2 (we all agree)

Q2: If there are two nodes, connected directly via half duplex, with the same wires, to a switch, and the switch is powered down, how many collision domains are there?  A: 0 (all should agree)

However, by analogy concerning many posts, some would argue Q2.  They would say, but if we power up the switch, there are 2 CD.  So what, that's not the specific assumptions for Q2.

Q3: If there are two nodes, connected directly via full duplex, with different wires, and actually using the full duplex to a switch, and the switch is powered on, how many collision domains are there?  The original had a typo and said 2.  This correction states A: 0 (all should agree)

By the same argument, if someone claims there are two collision domains, if it was running full duplex, my response is simply, that is not the correct assumption, in the specific question.

What is very interesting is that these disputes have been going on for ten years and we, the community, can still not come to any agreement :-)

[Joseph W. Doherty]

"Q2: If there are two nodes, connected directly via half duplex, with the same wires, to a switch, and the switch is powered down, how many collision domains are there? A: 0 (all should agree)

However, by analogy concerning many posts, some would argue Q2. They would say, but if we power up the switch, there are 2 CD. So what, that's not the specific assumptions for Q2."

Having a powered off switch sort of creates a "trick" question. By design, there's still two collision domains. By operation, with a powered off switch, there's no active collision domains. What point are you trying to make using a powered off switch? I.e. how does a powered off switch impact the original question or subsequent posts?

"Q3: If there are two nodes, connected directly via full duplex, with different wires, and actually using the full duplex to a switch, and the switch is powered on, how many collision domains are there? A: 2 (all should agree)"

If you're saying two hosts, each with its own connection to the same switch, on different switch ports, and the switch ports and host NICs are running full duplex, I don't agree there are two collision domains. If you say there are collision domains, please explain what frames might collide.

"What is very interesting is that these disputes have been going on for ten years and we, the community, can still not come to any agreement :-)"

For this type of question, I don't see it as disputable. This question, I believe, has a correct answer. However, what can cause confusion if the environment isn't fully defined. Perhaps that's the point you're trying to make?

That said, anyone with an answer should be able to explain how they came by it.

[Teacher Todd Booth]

@joseph> That said, anyone with an answer should be able to explain how they came by it.

That is the perfect solution.  I'll add that "should be able to explain", should require one to specify what their specific assumptions and/or conditions were, since there can be different correct answers, based on different assumptions and/or conditions.

So Joseph has done a great job!  We now can hopefully agree, that with regard to, for example, 5 entirely different assumptions and/or conditions, what the correct answer is, for each of these 5 very different cases.

Job well done Joseph!!!

[Joseph W. Doherty]

I 100% agree conditions and assumptions can certainly change an answer, and with different conditions and assumptions, different answers can be correct.

[Teacher Todd Booth]

There's an interesting way to look at the how many collision domains there are, on a full duplex link, with two sets of wires.

In the old days, we had a long shared coax Ethernet cable.  So there was one collision domain.

On a full duplex link, with two sets of wires, we actually have the following two (perhaps possible) collision domains:

Wire set 1: A sending to B

Wire set 2: B sending to A

Is Wire set 1, on its own, a collision domain?

Either we do, or we don't, have a collision domain on Wire set 1. ( 0 or 1 collision domains )

Whatever number of collision domains on Wire set 1, we have the same number on Wire set 2.

So on the cable as a whole (1+2), we have either 0 or 2 collision domains

Either way, we know that there is NOT 1 collision domain :-)