STP Limitations

[smitty6504]

STP is said to have a limit of 7 switches daisy chained together. Attached is a diagram of a network that has a fiber loop, does this meet the condition? Could you add one or two more switches in the fiber loop and still be inside the limitation?

[Nagaraja Thanthry]

Hello,

Your setup looks fine. If you are planning to add additional switches to the 6 switch loop, then you need to configure spanning-tree parameters manually on some of the switches such that the spanning-tree blocks a port in the middle of the loop.

On a different note, I understand that you need the loop to ensure redundant path. But what kind of switches are you using? If they are 3750's, then instead of creating a loop like this, you can configure stacking and configure cross-stack etherchannel. That would be a better solution as compared to spanning-tree loop. But it does have distance considerations (all switches need to be physically closer to each other).

Regards,

NT

[Peter Paluch]

Nagaraja,

It seems that we have both answered in a similar way Nevertheless, you have caught my attention with the following statement:

If you are planning to add additional switches to the 6 switch loop,
then you need to configure spanning-tree parameters manually on some of
the switches such that the spanning-tree blocks a port in the middle of
the loop.

Are you sure about this? I understand that it is recommended to increase the STP timers for networks larger in their diameter but even the formulas used to compute the default timers for STP are very conservative and provide quite a reasonable reserve so that even a larger network is not necessarily impaired with the default timers (the article at http://www.cisco.com/en/US/tech/tk389/tk621/technologies_tech_note09186a0080094954.shtml describes the computation amply but the estimations of various timers are outspokenly oversized).

Best regards,

Peter

[Nagaraja Thanthry]

Hello Peter,

Indeed we answered in the same way. The reason I talked about manually configuring the switches to modify the STP behavior is not related to timer but related to topology itself. What I meant was to change the priority of interfaces such that the spanning-tree breaks the loop somewhere in the middle so the loop looks like two branches with equal number of switches. So, from performance perspective, the packets need to traverse through fewer switches. If you let STP handle the loop and make a loop free topology, it could so happen that (depending upon the interfaces along the path) one of the branches has more switches than the other. In that case, the packets in the longer branch might suffer from performance perspective. Manually configuring the topology and knowing how the topology looks any time of the day always helps in troubleshooting as well.

Regards,

NT

[Peter Paluch]

Hello Nagaraja,

Oh, yes, I see your point now. You're right, of course. Thanks!

Best regards,

Peter

[Peter Paluch]

Hello,

The saying about maximum of 7 switched daisy-chained together is a myth. The STP is in fact limited to "max_age+1" switches linked in a row. However, the default timers for STP have been calculated with 7 daisy-chained switches so this is probably the source of this saying.

Nevertheless, I do not believe you will run into any issues if you enlarge your network a bit more. Take into consideration, however, that the current best design practices do not result in having 7 or more switches daisy-connected together, and you should consider modifying the topology of your network to decrease its diameter.

Also think of using RSTP that does not rely on timers for its correct operation.

Best regards,

Peter