Connecting switches in daisy chain / ring

Question

I have a network of 15 switches which are all connected together throughout a building (1 per floor) via daisy chain configuration. There is one cat-6 wire between each switch. All switches have 100mbits/s ports. Switch2-switch15 only have 1 device connected to it which is transmitting data back to a device on switch 1.

I have read that daisy chaining is not good for this amount of switches, so I want to add a home-run from the furthest switch to the first switch. The distance between the first & last is over 300ft, so I do not think it would work.

Is it possible to home-run the furthest switch to the middle switch and then from the middle switch back to the first?

Or maybe an even better idea? Thanks.

Answer 1

I'm going to go out on a limb here based on your username, the 100Mbps switches in 2021 and guess that you're looking at an industrial automation/building management network?

As Tuen mentioned, if these really are unmanaged switches, they will most likely NOT be running spanning-tree and any loop in the topology will break the entire network.

I would seriously consider replacing these devices with a managed industrial switch as these typically have a ring-redundancy protocol available (MRP being the most common/supported, followed by proprietary flavours like Hiperring, Turboring etc.) which will absolutely support this sort of topology with sub-50ms convergence time for dozens of devices in a ring.

Daisy-chain topologies like this are quite common and often the only viable way to provision networks in industrial environments - think Gas Pipelines or Railway corridors that need to collect telemetry data from mid-points that could be 100s of kms away from civilisation. Tree networks do not make sense in these environments and nor does Spanning-Tree.

If you replaced these devices with managed switches and deploy a ring-redundancy protocol like MRP or G.8032/ERPS, then to solve your specific problem with distance limitations, you could either complete the ring with a run of multi-mode optical fibre from Switch 1 to 15, or if you need to stick with copper, you could "braid" the connections by joining odd switch numbers on the way up the building and even numbered switches on the way down eg:

1-3-5-7-9-11-13-15-14-12-10-8-6-4-2-1

This way your copper runs would all be no more than 2 floors long.

Answer 2

Daisy chaining switches hurts latency as all switches' delays in a path add up. Performance doesn't seem to be your problem though as you're still on Fast Ethernet.

Chaining is also the worst topology reliabilty-wise. Any switch in a chain failing breaks the chain.

Deep (long) topologies also exclude redundant meshing using a spanning tree topology. RSTP/MSTP have a design limit of seven hops.

Ethernet is most comfortable in a tree - a center switch with all the other connected in one or at most two layers. If redundancy is required, use two center switches and two links to each of the others with spanning tree configured correctly (the center switches need to be root and standby root switches, with the lowest and second lowest bridge priority values).

Any form of ring or meshing requires a spanning-tree protocol or similar. Without that, the bridge loops causes a broadcast storm and brings down the network, usually in a matter of seconds. In other words, loops with unmanaged switches are fatal.

Using twisted-pair cabling, the longest legal link is 100 m (90 m solid-core + 10 m flexible patch cable). Longer links can be realized with fiber (multi-mode is good for a few 100 m, depending on cable grade and speed; single-mode runs for at least 10 km).

Is it possible to home-run the furthest switch to the middle switch and then from the middle switch back to the first?

Yes, that would be an improvement, but it requires STP.

If your cabling permits, connect all other switches to switch 8. If there's just a single host (or two), you can omit their switch and connect them directly, as long as switch 8's port count is sufficient. Use fiber for the runs longer than 100 m. Generally, try to keep the hop count to the center as short as possible - improving latency and removing single points of failure.

If you're restricted with budget, try to come closer to that goal. Depending on the cable lengths and count, you could e.g. connect switches 4-7 and 9-12 directly to 8, elect switches 4 and 12 as intermediate switches and connect 1-3 and 13-15 there.