- Posts: 13
- Thank you received: 0
Flooding incase of Switches..
- Datacom_guy
- Topic Author
- Offline
- New Member
Less
More
19 years 4 months ago #9220
by Datacom_guy
Flooding incase of Switches.. was created by Datacom_guy
When a switch does not know on which port to forward an incoming frame, it floods that frame on all the ports except the port from which it received that frame. In such case I was just wondering that if that destination node is made to send some sort of acknowledgment frame then the switch can learn the address of that node, so in the future the switch will just forward the frames to the correct port instead of flooding the frames everytime until it learns (which will happen when the destination node will transmit something!!). So if it happens that a node doesn't transmit any frame, the switch will not learn the port for that node and hence flood on all the ports...
Any comments on this are welcomed...
Any comments on this are welcomed...
19 years 4 months ago #9222
by Chris
Chris Partsenidis.
Founder & Editor-in-Chief
www.Firewall.cx
Replied by Chris on topic Re: Flooding incase of Switches..
DataCom,
Your correct in regards to the functionality of a switch - if the destination node's MAC Address is unknown, it will forward the frame to all ports.
When the destination node answers, the switch will note the port the answer came through and all future conversations will not be replicated to all ports.
However, I actually got confused as to if your asking a question or not; Can you please help clarify the purpose of the post as I've just returned from a long trip and are still a bit dizzy
Cheers,
Your correct in regards to the functionality of a switch - if the destination node's MAC Address is unknown, it will forward the frame to all ports.
When the destination node answers, the switch will note the port the answer came through and all future conversations will not be replicated to all ports.
However, I actually got confused as to if your asking a question or not; Can you please help clarify the purpose of the post as I've just returned from a long trip and are still a bit dizzy
Cheers,
Chris Partsenidis.
Founder & Editor-in-Chief
www.Firewall.cx
- Datacom_guy
- Topic Author
- Offline
- New Member
Less
More
- Posts: 13
- Thank you received: 0
19 years 3 months ago #9224
by Datacom_guy
Replied by Datacom_guy on topic Re: Flooding incase of Switches..
Chris,
I hope u've enjoyed u r vacation
Well while I was learning the switch functionality and this point struck me but I didn't find any documentation about the destination node explicitly sending an acknowledgement to the switch so that it learns about it, so I posted it here. Though I later on realised that though this will save the future flooding of frames to a particular node, we need to have some mechanism whereby the destination node sends an acknowledgement frame to only those frames which were flooded (which signify that the switch has not learnt about that node) else the node will end up acknowledging every frame it receives!! I feel this is the reason that a switch floods every frame until it learns. I am still wondering if this can be implemented and any other disadvantages of it?
Cheers,
Datacom
I hope u've enjoyed u r vacation
Well while I was learning the switch functionality and this point struck me but I didn't find any documentation about the destination node explicitly sending an acknowledgement to the switch so that it learns about it, so I posted it here. Though I later on realised that though this will save the future flooding of frames to a particular node, we need to have some mechanism whereby the destination node sends an acknowledgement frame to only those frames which were flooded (which signify that the switch has not learnt about that node) else the node will end up acknowledging every frame it receives!! I feel this is the reason that a switch floods every frame until it learns. I am still wondering if this can be implemented and any other disadvantages of it?
Cheers,
Datacom
19 years 3 months ago #9298
by tiamat
Replied by tiamat on topic Re: Flooding incase of Switches..
datacom,
I think you're trying to create something that isn't needed... you already understand that a switch will flood all ports when the destination address isn't known and that when (if) the destination host responds, the switch will cache it's address and know to send any further traffic to it through that same port which it got a response from. What else is there? Hosts already don't respond to flooded traffic if it's not destined for them. If the traffic IS for them, they respond.
Are you saying that network hosts should broadcast their mac address info so that switches can learn about them and build their tables without having to flood traffic to all switchports?
I think you're trying to create something that isn't needed... you already understand that a switch will flood all ports when the destination address isn't known and that when (if) the destination host responds, the switch will cache it's address and know to send any further traffic to it through that same port which it got a response from. What else is there? Hosts already don't respond to flooded traffic if it's not destined for them. If the traffic IS for them, they respond.
Are you saying that network hosts should broadcast their mac address info so that switches can learn about them and build their tables without having to flood traffic to all switchports?
19 years 3 months ago #9314
by Tarun
Next would be SP (Service Provider)
CCNA, CCNP (Switching), CCIE#20640
Replied by Tarun on topic Re: Flooding incase of Switches..
Well, i would disagree tiamat, datacom made a very good observation.
Let me try to make it a li'l more clearer as to what he is saying,
1) I have "Computer A" & "Computer B" connected to each other using a switch.
2) It is being assumed that the switch has just been plugged in so it does not has anything in its MAC table.
3) "Computer A" sends a UDP data packet to "Computer B".
4) Since the switch does not has any entry in the table, it just floods the packet on all the ports except for the one on which it received the data packet.
5) "Computer B" receives the data packet. ( But since it was a UDP packet does not sends any acknowldgement).
5) Now the switch knows where "Computer A" is on the network but it still does not knows where "Computer B" is because "Computer B" did not send any packet as yet to the switch & it won't as far as this communication is concerned.
Now what Datacom is trying to say is that if at this point "Computer B" after recognizing that this data packed was received because of flooding from the switch, instead of just reading the packet & sitting, it could have sent the acknowldgement back to the switch so that the switche's now knows the location of both "Computer A" & "Computer B". What this will do is to speed up the process of building up the MAC table on the switch & also reduce the traffic on the network
Let me know Datacom if i got it correct ?
Let me try to make it a li'l more clearer as to what he is saying,
1) I have "Computer A" & "Computer B" connected to each other using a switch.
2) It is being assumed that the switch has just been plugged in so it does not has anything in its MAC table.
3) "Computer A" sends a UDP data packet to "Computer B".
4) Since the switch does not has any entry in the table, it just floods the packet on all the ports except for the one on which it received the data packet.
5) "Computer B" receives the data packet. ( But since it was a UDP packet does not sends any acknowldgement).
5) Now the switch knows where "Computer A" is on the network but it still does not knows where "Computer B" is because "Computer B" did not send any packet as yet to the switch & it won't as far as this communication is concerned.
Now what Datacom is trying to say is that if at this point "Computer B" after recognizing that this data packed was received because of flooding from the switch, instead of just reading the packet & sitting, it could have sent the acknowldgement back to the switch so that the switche's now knows the location of both "Computer A" & "Computer B". What this will do is to speed up the process of building up the MAC table on the switch & also reduce the traffic on the network
Let me know Datacom if i got it correct ?
Next would be SP (Service Provider)
CCNA, CCNP (Switching), CCIE#20640
- cybersorcerer
- Offline
- Senior Member
Less
More
- Posts: 123
- Thank you received: 0
19 years 3 months ago #9316
by cybersorcerer
"He who breaks something to find out what it is, has left the path of wisdom."
Gandalf the Grey
Replied by cybersorcerer on topic Re: Flooding incase of Switches..
Remember that a simple network like that is not very realistic. Also, be aware that ethernet was not designed to provide acknowledgement and leaves those sort of functions to the upper layer protocols. In most cases, "computer b" will eventually respond to a request or request data long before a broadcast storm becomes a problem.
"He who breaks something to find out what it is, has left the path of wisdom."
Gandalf the Grey
Time to create page: 0.132 seconds