Address of a switch port - generic question

[hostettle]

Hi,

I hope my question is not difficult and sufficiently clear.

For management, a switch has 2 kinds of address.

- The address of the whole switch

- The individual address of each of its ports

Would you have pleased a case where the individual address of the switch port is included in the DA address of the MAC frame?

I think it's more common to have this address in the SA of the MAC frame => we have STP, MRP.

Best regards,

Michel          

[Peter Paluch · Accepted solution]

Hello Michel,

Cisco switches indeed have lots of MAC address. Basically, the switch as a unit has its own MAC address, sometimes called a base MAC address. This MAC address is used, for example, as the part of Bridge ID in STP. However, this base MAC address does not - to my best knowledge - ever appear as a source or destination MAC address in a frame.

This base MAC address is then used to derive a unique MAC address for individual ports on the switch, both physical and virtual (Port-channels, SVIs, etc.). These addresses may be used as source or destination MAC addresses in frames. For SVIs, the need for a MAC address is completely clear - a SVI must behave just like a router's interface with its own unique MAC address. MAC addresses on physical switchports are used as source MAC addresses in Layer2 management protocols spoken by the switch - STP, CDP, VTP, LACP, etc... These MAC addresses are also used as destination MAC addresses for LOOP frames. The other Layer2 management protocols mentioned before usually use multicast destination MAC addresses.

Does this clarify things a bit? Please feel welcome to ask further!

Best regards,

Peter

[hostettle]

Hi Peter,

Thanks for this reply.

> However, this base MAC address does not - to my best knowledge - ever

> appear as a source or destination MAC address in a frame.

It would be also my opinion

> This base MAC address is then used to derive a unique MAC address for

> individual ports on the switch, both physical and virtual (Port-channels, SVIs, etc.).

I thank it was the opposite, the base MAC address deduced from the address of the lowest number of the port

> a SVI must behave just like a router's interface with its own unique MAC address

Sorry for my ignorance. Could you give a few words for SVI frames. Is there an Ethertype?

> These MAC addresses are also used as destination MAC addresses for LOOP frames.

I'm interrested in these LOOP frames. Do you refer to OAM in the acess networks?

Thanks,

best regards,

Michel

[Peter Paluch · Accepted solution]

Hello Michel,

> This base MAC address is then used to derive a unique MAC address for 
> individual ports on the switch, both physical and virtual (Port-channels, SVIs, etc.).
I thank it was the opposite, the base MAC address deduced from the address of the lowest number of the port

Algorithmically, both approaches would produce the same results. However, it is easier for a switch management to define its base MAC address and the total number of MAC addresses this device may generate. The device then simply generates a MAC address for each of its ports.

> a SVI must behave just like a router's interface with its own unique MAC address
Sorry for my ignorance. Could you give a few words for SVI frames. Is there an Ethertype?

No apologies needed - on the contrary, I apologize for being unclear The SVI is an acronym for Switched Virtual Interface which is just a fancy name for interface Vlan X. The interface Vlan X for a particular VLAN X is an interface where the switch is assigned its own IP address, and also, if the switch has routing capabilities (i.e. is a multilayer switch), the IP address of this interface will be used by stations in the particular VLAN as the default gateway address. Naturally, when these stations send packets to other networks via this gateway, they will translate the gateway's IP address using ARP into the MAC address, which is one of the reasons why the SVI interface needs to have its own MAC address.

> These MAC addresses are also used as destination MAC addresses for LOOP frames.
I'm interrested in these LOOP frames. Do you refer to OAM in the acess networks?

No, this is something different (although the fact you know about OAM is admirable!). Actually, I suggest you read the following thread. It is quite lengthy but I guess it will answer your questions.

https://supportforums.cisco.com/thread/2001389

As usual - please feel welcome to ask further!

Best regards,

Peter

[hostettle]

Hi Peter,

> The SVI is an acronym for Switched Virtual Interface which is just a fancy name...

Many thanks for your patience and your time.

> Actually, I suggest you read the following thread. It is quite lengthy but I guess it will answer your questions.

I have to read again the thread.

It would seem that the loopback message, with Ethertype = 0x9000, is specific to Cisco could you confirm please?. I didn't see it in 802.1Q-2011, but I have to check again.

This LOOP frame tests the MAC connectivity (and not the PHY connectivity) between 2 opposite ports. It is a really good candidate to have SA = the MAC address of the source port, and DA = the MAC address of the sink port.

Best regards,

Michel

[Peter Paluch · Accepted solution]

Hello Michel,

It would seem that the loopback message, with Ethertype = 0x9000, is specific to Cisco could you confirm please?

Well, the LOOP frame was probably meant originally as an Ethernet version of PING, as Giuseppe explained in the thread I have linked. However, this functionality is to my best knowledge not implemented in any major operating system. Cisco appears to reuse these LOOP frames for a different purpose: each switch port sends a LOOP frame to itself (i.e. SMAC=DMAC). If there is no switching loop in the network, such a frame will never arrive back to the port that originated it. However, if there is a switching loop present in the network, this LOOP frame will be received by its originating port which will cause this port to immediately go to err-disabled state.

I didn't see it in 802.1Q-2011, but I have to check again.

I do not think it is described there. It was part of the original Ethernet II DIX specification but I do not believe it was retaken into IEEE Ethernet. In the thread I've linked, Giuseppe provided links to a PDF and a LWN discussion about the Ethernet Configuration Test Protocol suite which the LOOP message is a part of.

Best regards,

Peter

[hostettle]

Hi Peter,

> Cisco appears to reuse these LOOP frames for a different purpose: each

> switch port sends a LOOP frame to itself (i.e. SMAC=DMAC).

So, the LOOP frame is not a candidate to have the remote switch port adress in DA of the MAC frame.

Through that talking, I have another idea.

Suppose a switch with a function of IP routing. What is the address in the DA field if the IP datagram which is carried in the MAC frame has to be sent to the router part?

Is it the MAC switch port address or the base MAC address of the switch?

Thanks,

best regards,

Michel

[Peter Paluch · Accepted solution]

Hello Michel,

So, the LOOP frame is not a candidate to have the remote switch port adress in DA of the MAC frame.

No, it is not. In LOOP frames as used by Cisco, both the source and destination MAC address are set to the same address of the port that sent the frame.

Suppose a switch with a function of IP routing. What is the address  in the DA field if the IP datagram which is carried in the MAC frame has  to be sent to the router part?
Is it the MAC switch port address or the base MAC address of the switch?

Actually, it will be neither of these two. It will be the MAC address of the interface Vlan X interface (the SVI) for the VLAN in which the frame was received. The MAC address of a SVI is derived from the switch base MAC address but it is unique - i.e. not shared by any other switchport or other SVI.

Best regards,

Peter

[hostettle]

Hello Peter,

> The MAC address of a SVI is derived from the switch base MAC address but it is

> unique - i.e. not shared by any other switchport or other SVI.

Many thanks for these interresting information I've learned.

So, we don't have an example where the remote switchport address is inserted in the DA field of the MAC frame, in the LAN domain.

Thanks Peter,

best regards,

Michel

[Peter Paluch · Accepted solution]

Hello Michel,

So, we don't have an example where the remote switchport address is  inserted in the DA field of the MAC frame, in the LAN domain.

Quite correct. I do not know of any protocol that would use this kind of addressing.

It shall be said that the usage of MAC addresses as described in this thread is typical for Cisco Catalyst switches. Other vendors may use MAC addresses differently so I suggest caution when thinking about switches of different types or vendors.

You are very much welcome!

Best regards,

Peter

[hostettle]

Hello Peter,

I just want to add this.

In LAN we are not used to distinguish the user plane from the management plane.

I will suggest that the MAC address of a SVI is the switchport address in the user plane for a level 3 switch. This MAC address can be included in SA ou DA fields.

Now we have the MAC switchport address in the management plane. This MAC address is the "more often" used in SA field.

In WAN bridges, the management plane is more efficient, switchport addresses can be used in SA and DA fields.

Best regards,

Michel

[Peter Paluch · Accepted solution]

Hello Michel,

To be honest, I would not try to characterize MAC addresses and their usage based on management, control or data plane. You have to consider the fact that in our usual networks, we mostly have in-band control and management, meaning that the management of the switches and their control is done along with the data traffic, using the same ports, pathways and bandwidth. This is in quite a contrast to telecommunications equipment where management and control plane had a separate topology independent from data pathways.

In addition, different layer protocols would differ in their MAC address usage. Layer2 control protocols (STP, VTP, CDP, DTP, LLDP, PAgP, LACP, LOOP, ...) mostly use the switchport MAC addresses as sources. Layer3 control protocols (all routing protocols, first hop redundancy protocols like HSRP, VRRP, GLBP, Telnet/SSH/SNMP, ...) are sent from L3 interfaces which is mostly SVIs, so their messages would be ultimately encapsulated using SVI MAC addresses as sources. So the criterion of management/control vs. data plane traffic does not really define anything.

There is one thing I realized during our discussions: with the Layer2 control protocols mentioned above, it is actually necessary to use switchport MAC addresses as sources. Using the base MAC address of a switch would be bad: imagine STP BPDUs being sent from multiple ports towards a single neighboring switch (assume they're interconnected with multiple links). If the base MAC address was used as a source MAC address in all outgoing BPDUs, the opposite switch would complain about the same source MAC address flapping between its own ports which would be true. It is in fact necessary that for Layer2 control protocols that are used by switches to communicate among themselves, the messages sourced by individual switchports must be sourced from unique MAC addresses.

I am not sure if this helps... it's quite convoluted.

Best regards,

Peter

[hostettle]

Hello Peter,

> ...the management of the switches and their control is done along with the data traffic,

> using the same ports, pathways and bandwidth.

It's a characteritic of Ethernet 802.3 (MAC over PHY) to use all the LAN network for all the planes, it's a connectionless transmission. This is in contrast with other Ethernet architectures as Ethernet over SDH (MAC over SDH, or MAC over MPLS over SDH), these architectures are oriented connection.

But... if the transmission is done horizontally in all the LAN network (for Ethernet 802.3), the planes are distinguished inside the equipements, vertically, at the level 2, 3 or at TCP/IP application.

SNMP has actions in the MAC management plane, for the configuration or the performance (and perhaps in the control plane, I don't remember).

STP has actions only in the MAC control plane, to prevent loop and to provide survivability. It can modify the conversational context (the connectivity organisation).

We are far from the switchport subject. But, we don't confuse the physical ports (the PHY ports) from the logical ports at level 2 or higher.

Best regards,

Michel

[Peter Paluch · Accepted solution]

Hello Michel,

I apologize... while I agree with most of what you wrote, I am not quite sure where you're heading. I do not entirely understand the meaning and the context. Perhaps if you could clarify it a little bit... I am really sorry for not being able to understand better.

Best regards,

Peter

[hostettle]

Hello Peter,

It's to me to apologize. I'm just an amateur, a little bit interested by the many-faceted Ethernet, the concepts from LAN to MEF. With pleasure, we can discuss that through future threads. I appreciated your writing.

In the originating subject I permit me to ask you this question.

What is the relation between the address of the management port and the base MAC address of the switch?

Thanks,

best regards,

Michel