I did many labs and read some texts on Spanning-Tree. I wonder everytime what role timers play in RSTP (802.1W protocol) ?
- Listening delay : My question is not about that delay. I did understand that it existed in 802.1D, but did no longer exists in 802.1W.
- Learning delay : That delay is still 15 seconds. But, Wendell Odom (to name the CISCO official CCNA book) says : "RSTP convergence times are typically less than 10 seconds". How come ? Do Switch ports not wait until learning delay expires when switch is powered on or switch port is moving from "Blocking" to "Forwarding" state ?
- MaxAge timer : Is really that timer needed in RSTP ? I understood that RSTP convergence starts as soon as witch receives inferior BPDU. If the reason that timer has not been removed is for keeping compatibility with 802.1D, why his value has been changed ?
Thanks in advance for any help !
RSTP is not timer based. However, if a a bridge running 802.1w needs to interoperate with a bridge running 802.1d, I think the timers would be necessary for backwards compatibiltiy reasons.
default cisco switches are in pvst we need to change it to Rpvst+
so see timers in STP
1.BLOCKING 20 SECONDS
Receives bpdus not learns MAC address nor sending data from computers
2.LISTENING 15 SECONDS
it will receive and send bpdus still not learning mac or data transmission
3.LEARNING 15 SECONDS
sending and receiving bpdus note in this state it learns mac address but still no data transmission......
finall.y we can start sending and receiving data
see in RSTP+
there is no state like blocking and listening only learning and forwarding (discarding) so it was much faster than pvst but there is no timer for RST+ RSTP is not timer based
Thanks Hari for the quick and clear wrap-up on STP timers.
I still have some curious questions :
1) About the MaxAge timer ; CCNA books tell that in RSTP it has been reduced to 6 seconds. If that timer is kept in RSTP just for compatibility with 802.1D, why did it not be kept to 20 seconds ?
2) CCNA books tell also that Listening timer does no longer exist in RSTP ; but, Learning timer is still there. ???
3) Even if many mechanisms are built-in in the protocol for making it faster. When the switch is restarted, I think ports will have to transition on Learning state before getting Forwarding state. This process seems very faster on RSTP ; but, unfortunately, the only Learning delay that has been documented for the two protocols (STP and RSTP) is 15 seconds.
Please, find attached to the post what Wendell Odom says about RSTP timers. Read "Link-Type Point-to-Point" section.
The forward delay on pure 802.1w/RSTP enviroments is not used. The forward delay timer is not used because of the synchronization process, but remember that the max age timer is certainly used by 802.1w/RSTP, but insted of being 20 seconds is basically 6 seconds (three missed BPDUs in a row). 802.1w/RSTP is an IEEE standard and it must be backward compatible with 802.1D/STP and for this reason a switch running 802.1w/RSTP can run 802.1D/STP on a per port-basis and apply the 802.1D controls in the specific port. This is detected thanks to the protocol version inside the BPDUs, which is number 2.
Thanks Elvin for the information about the BPDU version. By the way, a quick question : Are you telling me that I can have (in the same LAN) some switches behaving like STP and some switches behaving like RSTP ?
About the page I sent. I justed wanted to show that Wendell Odom did not say that "MaxAge timer is used solely for compatible purposes of 802.1W with 802.1D". My understanding of what Wendell Odom is saying is that : detection of indirect link failure could take up to 6 seconds.
If you have three switches redundantly interconnected you can run 802.1w/RSTP specifically with the switches that supports it and 802.1D/STP with the 802.1D-only speaker this is basically the reason why 802.1w/RSTP is backwards compatible with 802.1D/STP.
Yes, one of the uses for the max age timer is for indirect failures. Note that the 6 seconds timer will only work for 802.1w/RSTP interconnected switches. If a 802.1w/RSTP switch is connected with other switch that works with 802.1D-only it will set it's parameters to this version (the 802.1D old STP), so it will be able to use the 15 seconds forward delay and 20 seconds max age timer.
If a 802.1w/RSTP switch is connected with other switch that works with 802.1D-only it will set it's parameters to this version (the 802.1D old STP), so it will be able to use the 15 seconds forward delay and 20 seconds max age timer.
According to what you are saying ; if I have a LAN with 9 switches running 802.1W. If the 9 switches are connected to 1 switch running 802.1D ; then, all 10 switches will have 802.1D behavior ?
Note that the 6 seconds timer will only work for 802.1w/RSTP interconnected switches.
Then ; in a RSTP (802.1W) deployment, convergence delay (after failure) could be up to 6 seconds ?
Not all 9 switches, but instead the switch (es) directly connected to the 802.1D-speaker/switch. It is what i decribed earlier about the "per-port behavior".On the second part the 802.1w/RSTP max age is 6 seconds, but it is only used for indirect failures. Remember that the reconverge on 802.1w can be even at the milisecond level, but there are certain events where the switches need to use the max age, so the standard maximun time for convergence when the max age timer is used is about (theorically speaking) 7 seconds.
Not all 9 switches, but instead the switch (es) directly connected to the 802.1D-speaker/switch. It is what i decribed earlier about the "per-port behavior".
Im not talking about "disabling" STP on a per-port basis, but instead being compatible with older STP versions. As a side note STP can be disable on a per-port basis. Look for the BPDU filter feature.