Question about Spanning Tree Protocol

Chuu

I have some questions about STP.

1. I want to make 100% sure I am understanding this correctly, but STP doesn't actually form a tree. It forms an a-cyclic graph (assume you cannot follow a path that you took to reach the current node). True or False?

2. This is what confuses me. It was briefly mentioned that a downside to not using STP is that messages get duplicated. The problem is, let's say we have the following switch network (this is verbatum from odom) after STP has negotiated the routes.

S1 : fa0 -> S2, fa1->S3
S2 : fa0 -> S1, fa1 ->S3
S3 : fa0 -> S1, fa1 -> S2

Basically it's just three switches in a triangle. Now STP establishes what ports block and forward, and we get this:

S1 (root) : fa0 [forward], fa1 [forward]
S2 : fa0 [forward], fa2 [forward]
S3 : fa0 [forward], fa2 [block]

The thing is, even though broadcasts do "die," the root node is going to get 2 copies of the broadcast if S2 is the originiator, one via S2_fa0, and one via S2_fa1->S3_fa0.

So I want to make 100% sure nothing is going wrong here, i.e. even though we're using STP, the only gaurentee is that every frame has a finite lifetime i.e. is eventually consumed by a switch/client, not that each client is guaranteed received frames are unduplicated.

dynamik

Chuu wrote:

The thing is, even though broadcasts do "die," the root node is going to get 2 copies of the broadcast if S2 is the originiator, one via S2_fa0, and one via S2_fa1->S3_fa0.

Do you mean if you're not using STP? If the port on S3 that goes to S1 is blocking, it's not going to forward anything to that switch. S1 will only receive the broadcast from S2.

Chuu wrote:

So I want to make 100% sure nothing is going wrong here, i.e. even though we're using STP, the only gaurentee is that every frame has a finite lifetime i.e. is eventually consumed by a switch/client, not that each client is guaranteed received frames are unduplicated.

I'm not sure what you mean by finite lifetime and dying. The broadcast will be forwarded out of every port except the one it came in on. If you have multiple connections and aren't using STP (or one of the variants), that's going to lead to a broadcast storm that isn't going to end until you break the loop. The broadcasts aren't simply going to time-out or anything like that.

yuriz43

1 (root) : fa0 [forward], fa1 [forward]
S2 : fa0 [forward], fa2 [forward]
S3 : fa0 [forward], fa2 [block]




S3 will not receive the frame from S2, it's Fa2 interface is in a blocking state. The only thing S3 will let through on its Fa2 interface is BPDUs.

Therefore, when a system connected to S2 sends a broadcast, the frames will be sent out all of its interfaces because one of its interfaces is a root port, and the other is a DP, both being in a forwarding state.( of course S3 will block the frames received directly from S2) Once switch1 receives the broadcast it will send it out all of its ports (except the one it originated from) and this is how S3 will receive the frame.( through its root port to S1).[/quote]

dynamik

So where does fa2 fit into the equation? I just thought that was a typo since you didn't mention it earlier...

Chuu

Yeah, there's a typo in there. Really I need a picture to show what I'm asking, hopefully I'll scan it in tomorrow.

I understand what STP does and how it does once the blocking/forwarding states are "stable." It's just the learning algorithm is confusing me a bit, at least the way Odom describes.

kryolla

Odom usually does a god job in explaining things. If you need a more visual presentation try to do a search.

dynamik

Check out the spanning tree sheet here: http://packetlife.net/cheatsheets/