Thanks a lot Francois!

hi Francois!

Just want to clearfy one thing.

In traditional stp, root bridge generates bpdu while other switches act as relay.

for example:

root Sw1D-------------- R-sw2D---------R-sw3

where:

D stands for designated port

R stands for root port

When sw2 receives bpdu from sw1, it modifies sender id among other fields and relays it on its designated port.

According to cisco document related to rstp"Bpdu are sent every hello-time and not simply relayed any more. With 802.1D, a non-root bridge only generates bpdu when it receives one on the port"

According to above, if sw2 does not receive bpdu , it should not be able to generate one.

Am i right?

If i am right, that means sw3 should not be receiving inferior bpdu from sw2 ifsw2 does not receive bpdu from sw1.

thanks a lot!

Hi Sarah,

That's correct. With STP, BPDUs are relayed from the root toward the edges. That's why the aging time is so large. It must account for the potential loss of the BPDU anywhere on the path from the root to a given bridge that can be 6 hops away from the root. An RSTP bridge keeps transmitting the root information every hello-time, which means that if its neighbor does not receive it, there has been a communication problem on the last link. That's why the aging out can be much faster.

Anyway, in your example, sw2 relay the root information to sw3. If it stop receiving this information, it will indeed stop sending BPDUs to sw3 but it will also eventually age out the root information. Then it will declare itself root and start sending BPDUs again (its own BPDUs this time) to sw3.

Regards,

Francois

thanks a lot Francois! Your reply make me question the justification of back bone feature.

My argument is based on" A switch will only generate inferior bpdu on designated port once the max age timer expires on the root port without receiving any bpdu from root"

We ignore propagation delay, proccesing delay for bpdu for our example.

sw1 and sw2 are connected to switched network by root port f0/1 on each switch ( sw1 and sw2)

sw1 and sw2 are also connected to each other by f0/2 ports

sw2 f0/2 port is designated port while sw1 f0/2 is blocked port. Back bone feature is enabled for the network

sw2 receives bpdu from root at t1

max-age timer is reset

Since we ignore the propagation delay processing delay for bpdu by sw2 for simplicity, that means sw1 receives bpdu at the same time

sw1 receives bpdu on f0/2 from sw2 at t1

max-age timer is reset

sw2 stop receiving bpdu from root, sw2 waits for t1+20 seconds for max -age timer to expires.similarly sw1 waits for max-age timer to expire i.e it will wait for t1+20 seconds

after t1+20 sec, max -age timer expires, sw2 sends inferior bpdu to sw1

sw1 accepts this bpdu as it max-age timer has already expired and put blocked port in listening state.

Even if take into account the variable we ignore such as propagation delay, processing delay, there should not be any difference of more than few seconds between the max-age timer on sw2 root port and max-age timer on sw1 blocked port which in turn determines when sw2 can generate bpdu and when sw1 can accept the inferior bpdu from sw2.

My text book says backbone feature help save 20 second( by doing away with max-age time), i feel it saves just few seconds

Any comment will be greatly appreciated.

thanks a lot!

I'm not a big fan of backbone fast as converging in 30 seconds instead of 50 does not seem to deserve running this additional feature on all your backbone switches. However, it introduced a very important behavior that was included in RSTP: the ability to act on inferior bpdus. In the scenario you described, backbonefast would be of no help because sw2 just stop receiving bpdus. Usually, when there is a failure, a link down is detected. For instance, if sw2 detects its root port going down, it can immedaitely age out its root information. Now, it only has a designated port left, so it assumes it is the root and start sending worse bpdus to sw1. Sw1 receives worse bpdus on its alternate port f0/2 and just *ignore* them. It's because it has not yet aged out the alternate information it used to receive from sw2. That's were the 20 seconds are wasted. Sw2 tell sw1 that it has lost its connection to the root, and sw1 ignores this notification. Backbone fast accepts this inferior information and allows sw2 to immediately age out the alternate information it had received on f0/2. As a result, this f0/2 becomes a designated port, and transition to forwarding in 30 seconds. This would have taken 50 seconds without backbone fast.

RSTP behave the same, except that the designated port on sw1 proposes, receives an agreement from sw2 and then goes forwarding.

Regards,

Francois

Thanks a lot Francois! thanks every body who are helping me understand cisco technology.

Hi Francois!

I just want to ask if a switch experiences a direct link failure(root port), will it send bpdu claiming to be root on blocked port? thanks a lot!

thanks for your reply Francois!

please excuse my long-winded questions.

so if we have three switches sw1,sw2,sw3, connected in traingular fashion, with sw1 as root.

sw2 and sw3 have f1 as root port.

let focus on the segment which connect sw2 and sw3.

sw2 sw3

f2(alt) f2(D)f3(backup)

-----------------------

if sw3 loses root port f1,will it transmit bpdu on back up port?

you said" The port will stay back up port however,until new BPDUs are received on it, and it will not transmit any BPDU in the meantime"

My question how can backup port receive new bpdu on the segment since alternate port does not tranmit bpdu?

thanks a lot!