STP and LAN segments

Rearden

I'm confused about the designated switch concept of spanning tree protocol and I'm pretty sure that it's a result of confusion on what exactly a 'lan segment' is. So, Here's a diagram from 3com that i'd like to ask some questions about:




Okay, so in this diagram, we have three switches, er. . .bridges. So, if this LAN used hubs, the definition of a segment looks like it would be all of the devices that can talk to each other without going through a bridge? Correct?


Okay, let's say that those bridges are switches. How do things work now? The best thing that I can come up with is that these are distribution level switches that have multiple connections to an access level switch? If this is the case, do the boundraies for the segments stay the same?


I suppose that in the end it all boils down to: what exactly is a segment and how does the definition change when moving to a switched network?

loboernesto

think of a segment as a colission domain
segmentation increases collision domains.

cheers

dissolved

Rearden wrote:

I'm confused about the designated switch concept of spanning tree protocol and I'm pretty sure that it's a result of confusion on what exactly a 'lan segment' is. So, Here's a diagram from 3com that i'd like to ask some questions about:




Okay, so in this diagram, we have three switches, er. . .bridges. So, if this LAN used hubs, the definition of a segment looks like it would be all of the devices that can talk to each other without going through a bridge? Correct?

No. Hosts in the same segment speak to each other without the need for a layer 3 device. They address each other by MAC address

When a host on segment 1 wants to communicate with a host on segment 2, they will have to send their packets to a layer 3 device to get to the other segment

Rearden wrote:

Okay, let's say that those bridges are switches. How do things work now? The best thing that I can come up with is that these are distribution level switches that have multiple connections to an access level switch? If this is the case, do the boundraies for the segments stay the same?

Not sure what youre asking. Switches speak to each other via BPDU packets. They are sent every few seconds. The STP protocol carries the BPDU packets. Think of bridges like switches, without the same number of ports. The purpose of STP is to have only one link between switches. All redudant physical links will be disabled, until of course they are called upon. I do not think BPDU messages leave a network segment. Not positive on this though

Rearden wrote:

I suppose that in the end it all boils down to: what exactly is a segment and how does the definition change when moving to a switched network?

A segment is a network segment, what makes segments is layer 3 devices. An example of a network segment is hosts that all have 192.168.3.0 255.255.255.0 addresses. Another segment may be 192.168.4.0 255.255.255.0

loboernesto

I don't agree with all that dissolved said.. we may be talking about different things.

If you are confused about stp you have to focus on the ethernet segmets (to be more specific). An ethernet segment is defined by the collision domain, a layer 3 device will define a segment but also a layer 2 device will as is the case of a bridge or a switch.
In your graphic you can see that each port in a bridge is a different segment, that is because the bridge makes a differnet collision domain on each port. In a switch every single port will be a different ethernet segment. Think around that and you will understand STP

Rearden

So is a segment a layer 2 concept?

loboernesto

I don't know if you can define segment as layer 2 or layer 3...
I'll quote some of the first paragraph of Wendell Odom's cert guide (STP chapter):

"To prevent looping frames, STP blocks some ports from forwarding frames so that only one active path exists between any pair of LAN segments (collision domains)."

To my understanding a segment is collision domain and a network is broadcast domain.
cheers

dissolved

loboernesto wrote:

I don't agree with all that dissolved said.. we may be talking about different things.

If you are confused about stp you have to focus on the ethernet segmets (to be more specific). An ethernet segment is defined by the collision domain, a layer 3 device will define a segment but also a layer 2 device will as is the case of a bridge or a switch.
In your graphic you can see that each port in a bridge is a different segment, that is because the bridge makes a differnet collision domain on each port. In a switch every single port will be a different ethernet segment. Think around that and you will understand STP

What part do you disagree about? Network segment? A network segment is not defined by collision domains, it is defined by broadcast domains.
For a very trivial example-every switch on a port represents a different collision domain. Every port on a router represents a different broadcast domain.

The original posters diagram is a bad one, but what do you expect from 3com. A lan segment is a bad word to use in that diagram. From that diagram, one would be lead to believe VLANs are in use. If they were smart, they would have used hosts connected to the switch. Not "lan segments"

Here's a good diagram

bmauro

I'm also kinda confused on what is being asked.

For STP - a LAN Segment is what loboernesto posted. For STP each bridge (switch) has a root port that forwards, and for each LAN Segment a designated port is also forwarding.

So in regards to Spanning Tree - each link between two switches or bridges is considered a LAN Segment and the lowest cost port for that segment will be placed in a forwarding state.

Rearden

Ok, so in the new diagram, assume that switch 16 is the root, for no other reason than it's in the middle.

So, every port on root 16 is forwarding. . . the ports on the other switches connected to the root switch are forwarding. . . and which other ones?

I'm sorry I'm still confused on this I just don't really get what the designated switch/port is all about.

Webmaster

dissolved wrote:

loboernesto wrote:

I don't agree with all that dissolved said.. we may be talking about different things.

If you are confused about stp you have to focus on the ethernet segmets (to be more specific). An ethernet segment is defined by the collision domain, a layer 3 device will define a segment but also a layer 2 device will as is the case of a bridge or a switch.
In your graphic you can see that each port in a bridge is a different segment, that is because the bridge makes a differnet collision domain on each port. In a switch every single port will be a different ethernet segment. Think around that and you will understand STP

What part do you disagree about? Network segment? A network segment is not defined by collision domains, it is defined by broadcast domains.

loboernesto is correct. STP was originally developed for bridges, which segment a LAN in separate collision domains, not broadcast domains. The term segment can be used to indicate a portion of the network, which can be one or more collision domains or broadcast domains. In case of the 3com diagram, which is perfectly fine and a classic one used to explain STP, a segment = collision domain. All the LAN segments in the diagram (and possible hosts in it) belong to one and the same broadcast domain.

dissolved wrote:

No. Hosts in the same segment speak to each other without the need for a layer 3 device. They address each other by MAC address.

When a host on segment 1 wants to communicate with a host on segment 2, they will have to send their packets to a layer 3 device to get to the other segment

No, a LAN segment, as shown in the 3com diagram, is not a broadcast domain, the LAN is the broadcast domain, which is segmented into separate collision domains.

Segment a LAN and you get LAN segments. If a host in one LAN wants to communicate with a host in another LAN (where both LANs are in the same WAN/internetwork, hence use a different subnet, hence are in a different broadcast domain), 'then' you would need a layer 3 device. But hosts in the same LAN (but in different LAN segments), don't need a layer 3 device. Obviously the 3com diagram shows the LAN segments are connected by bridges, they provide communication between the LAN segments and layer 3 devices are not needed.

The purpose of STP is to have only one link between switches

That's a result of the purpose. The purpose of STP is to prevent loops (and the resulting broadcast storms and messed up MAC tables) by allowing only one active path to a LAN segment. So broadcasts (layer 2 broadcasts, to ff.ff.ff.ff.ff.ff) and unicast with an unknown address a switch receives on one port are not put back on the LAN segment by another switch that is connected (even indirectly) to the first switch. The designated port is that one interface that is allowed to send and receive on the LAN segment.

So if you would have two switches, with two links between them, you have 2 collision domains, 2 LAN segments, 1 LAN, 1 broadcast domain. Obviously, each switch only has one 'link' to a segment, but without STP, each interface would have a 'path' to both segments, creating a loop. So the STP example with 2 links between two switches, is just that, an example.

How LAN Switches Work

Understanding and Configuring Spanning Tree Protocol (STP)

Rearden

Alright, I think I have it figured out:

say you have this crude diagram: Assume that the core is the root. and switch 1 is the one on top of that 'stack'



The switches are arranged in a stack like manner. So . . ports and 51 and 52 on switch 1 are from the core and ports 48 and 49 are to switch 2 and are not set up as a channel-group. So, either port 48 or port 49 on switch 1 is the designated port to switch 2 and either port 51 or 52 is the designated root port.

So I think I finally understand what the idea of a designated port is. . so is switch 1 a designated switch from the root to switch 2?

masterk

Sorry to resurrect an old thread,

So above when you say Two switches are connected to each other with two links, and thats two "Collision domains". Do the collision domains refer to the whole switch in general? Because a switch is multiple collision domains?

Its like Wendel Odom's quote further up which i think more refers to the old days of hubs and bridges?

networker050184

Each port on a switch creates a separate collision domain.

masterk

Thanks Networker i have gathered that but, see this quote from above:

"To prevent looping frames, STP blocks some ports from forwarding frames so that only one active path exists between any pair of LAN segments (collision domains)."

How does that quote apply to linking two switches together? It applies to linking two or more hubs together, but not switches? As like you say every port in a switch is a collision domain, not the whole switch.

networker050184

masterk wrote: »

Thanks Networker i have gathered that but, see this quote from above:

"To prevent looping frames, STP blocks some ports from forwarding frames so that only one active path exists between any pair of LAN segments (collision domains)."

How does that quote apply to linking two switches together? It applies to linking two or more hubs together, but not switches? As like you say every port in a switch is a collision domain, not the whole switch.

It just means there will only be one path through the switched network to reach any given segment.

Forsaken_GA

Edit: ah, nevermind, didn't realize this was a lazarus thread!