|
Posted by Andrey Tarasov on June 26, 2008, 3:15 pm
If you were Registered and logged in, you could reply and use other advanced thread options
>> aaabb...@hotmail.com wrote:
>>> stp convergence time =50s or 30-50s? why?
>> For regular STP with default timers on Cisco equipment and properly
>> configured ports - it is 30 seconds.
>>
>> Regards,
>> Andrey.
>
> Hmmm. I would disagree with that.
That's fine. How about some facts?
> I too was somewhat confused by observing that re-convergence
> can take either about 30 seconds or about 45 seconds (maybe 35 and
> 50?).
>
> I believe that it is like this.
>
> For a port that is not connected that gets connected STP has to go
> through
> all of the states
>
> Blocking
> Listening
> Learning
> Forwarding
> 45-50 seconds - If I recall correctly
Why? 2 x forwarding delay is 28.5 sec. Where are additional 15-20 sec
coming from?
>
> For a port that is already in the network and is Blocking
> The transition to Forwarding is about 30-35 seconds
So what's the difference between newly connected vs already connected?
> The default timers will be easy to find on the web.
>
> Note that the times do not really range over the interval indicated
> but I have specified a range since I cannot remember the correct
> times and BPDUs are sent only every two seconds so there will
> be a range of times introduced by the port state change possibly
> occurring just before or just after a BPDU tx or rx event.
>
> Of course there are other considerations.
>
> For a newly connected port then there may be
> physical speed/duplex negotiation
> etherchannel negotiation
> trunking negotiation
Not relevant to original question. While etherchannel/trunking
negotiation indeed adds time, it's not part of STP convergence.
>
> There are also various STP optimisations
> portfast
That's enabled on edge ports only and doesn't affect convergence time at
all.
> uplink fast
> backbone fast
Cisco extensions and not part of original 802.1D spec.
> Lastly of course there is Rapid STP which is substantially changed
> from
> traditional STP and has much lower convergence times. I have not
> worked
> with it but have the idea that convergence times of one to a few
> seconds are
> reasonable.
Again, Rapid STP is completely different beast.
Regards,
Andrey.
| 
Yahoo! 
Google 
Windows Live 
del.icio.us 
digg 
Netscape 
XML site map