1

I'm designing a network which is physically arranged like:

|---------------------------[router]---------------------------|
|                                                              |
|--[s]--[s]--[s]--[s]--[s]--[s]  [s]--[s]--[s]--[s]--[s]--[s]--|

There's actually around 10 daisy chained switches [s] on each of those arms, each with 6 devices. Notice there's a gap in the middle of the bottom line.

The network is quite low traffic (about 10KB/s per device) for controlling motors in an art installation.

Since the daisy chains are very long, i'd love to be able to link together that gap in the middle:

|---------------------------[router]---------------------------|
|                                                              |
|--[s]--[s]--[s]--[s]--[s]--[s]--[s]--[s]--[s]--[s]--[s]--[s]--|

This would mean that if one of the links failed, we'd still have a connection everywhere. But in normal operation we would get a packet storm.

I'm wondering if it's possible to do something like:

                             [router]
                                 |
|----------------------[expensive switch]----------------------|
|                                                              |
|--[s]--[s]--[s]--[s]--[s]--[s]--[s]--[s]--[s]--[s]--[s]--[s]--|

Where the 'expensive switch' would stop the packet storm from happening because it would notice the packets going around in circles and it would have special features to stop this perpetuating. And by expensive, I'm still thinking

Note that it's not an option to change the cheap [s] switches. These are Cisco SF95D-08-GB unmanaged switches (i'd be interested if there is a significantly better option for the same price or less, but that would be a different question).

  • Did any answer help you? If so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you can provide and accept your own answer. – Ron Maupin Feb 19 '18 at 16:59
4

As long as your "expensive" switch runs Spanning Tree Protocol, you can connect your switches in a ring. STP will prevent packet storms by only allowing packets out one of its (ring) ports. But if there is a break somewhere, it will forward packets out both ports, so all your unmanaged switches will still receive frames.

Product recommendations are off-topic here, but there are many low cost switches from several manufacturers that would work here. For best results, make sure they support Rapid Spanning Tree (RSTP) 802.1w. Virtually all modern switches do.

EDIT: Note this will only work if the cheaper switches are not 802.1D compliant. If they are, then they will drop the BPDU data in the I2 frames, which means that the 'expensive' managed switch will not be able to correctly map the network.

  • Man typical i was writing somewhat the same. :-) +1 for sure. – user36472 Jul 29 '17 at 11:33
2

That won't work. The "expensive" switch (presumably able to build a spanning tree) needs to be aware of the network topology. For that it requires the others (contributing to the bridge loop) to participate in the spanning tree protocol.

Maybe you can run a few more cables, build the chain with fewer switches and then slightly more expensive ones become an option.

In addition to STP, some switches can also send loop probing packets. When these return to the switch that send it out it switches off(!) the receiving port for some time. This crude method could actually work in your scenario, limiting the extend of the broadcast storm - however, the "perpetrator" can't be activated as soon as the rest chain is interupted (the switch doesn't even have a way to tell), and will stay off for the configured time. Then, when the port's activated again the chain will be restored.

  • Why wouldn't a managed switch work? Spanning tree would block one port, unless there was a break in the ring. – Ron Trunk Jul 29 '17 at 10:54
  • Well, the simple switches shouldn't forward the BPDUs. If they do, it would work. – Zac67 Jul 29 '17 at 10:58
  • 1
    There's nothing magical about BPDUs. They're just l2 frames. Unmanned switches just forward them like any other. – Ron Trunk Jul 29 '17 at 11:07
  • 1
    802.1D compliant switches are supposed to drop them AFAIK. – Zac67 Jul 29 '17 at 12:31
  • so the conclusion is that as long as the unmanaged switches are not 802.1D compliant, then i should be able to get away with using a single STP-aware switch as my 'expensive switch'? – Elliot Woods Jul 30 '17 at 11:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.