I am building a network, with each node connected to each other using fiber media converters. They are actually a bridge with 2 ports: one Ethernet and one optical (or one single-mode and one multi-mode). I literally add 2 extra bridges between 2 nodes. For a tree topology, the number of bridges/switches between 2 far away devices is dramatically increased.

image for illustration purpose

What is the effect of having too many bridges/switches (or routers) between 2 devices? Is there any maximum number of bridges or routers I can have before the network becomes unreliable?

I am now thinking of a contrived example where I use multiple switches and routers to connect 2 devices which are 1km away, using 10x 100m twisted pair cable. Will it work?

  • One significant effect is that you become poorer. – abligh May 2 at 19:39

You might want to consider SFP modules/transceivers integrated into the switches instead of unmanaged converters - SFP modules in managed switches are much better to monitor and troubleshoot.

In your diagram, you connect the converters to the switches using multi-mode fiber. If those switches have SFP slots, just use single-mode transceivers instead and leave out the media converters. Make sure your buying compatible SFP modules, many switch vendors try to lock you in on their "original" modules.

Most media converters are no real bridges (which selectively forward by MAC address) but dual-simplex repeaters - these are capable of full-duplex transmission, yet they forward everything they receive. This doesn't really matter when they're connected to switch ports though.

Having multiple switches or converters in between two normal switches isn't really a problem (technically you can chain as many as you like) but it can make it harder to locate a problem when the link fails. Also, a larger number of devices fails more easily than a smaller number. I'd seriously recommend SFP modules in managed switches.

Twisted pair has a maximum reach of 100m (90m solid-core and 2x 5m stranded cable), so you'd need at least nine switches for regeneration distributed over the distance - that is usually very cumbersome when a problem occurs. With unmanaged switches, you'd have to literally check each switch and the links physically, in person. With managed switches, you'd still have more hardware to monitor and manage than with a single, longer-reach link.

Using single-mode fiber (SMF) and a single pair of 1000BASE-LX tranceivers will provide for better and more reliable service. 100BASE-FX on multi-mode fiber (MMF) would also make the distance, but likely won't be cheaper and MMF won't work for a gigabit upgrade. SMF is good even for 10GBASE-LR and beyond.

If the link is critical you should consider providing some redundancy. With managed switches you could aggregate two 1000BASE-LX links, so the links become redundant. If you'd like redundant switches you can run two links, each terminated on a different pair of switches, and use a spanning-tree protocol (RSTP or MSTP) to resolve the otherwise present bridge loop.

| improve this answer | |
  • From what I've heard, if a link cannot operate at 1000mbps, it will negotiate with the other end to fallback to 100mbps, and then 10mbps. Is the link speed a guaranteed constant value? I mean if it is 1gbps, then it shall remain 1gbps regardless of signal loss due to cable length and interference? I always thought that 1gbps is the maximum, and can be reduced (to 800-900mbps) by other factors. – Livy May 2 at 13:50
  • I was not very confident of chaining too many bridges/switches because of that reason. – Livy May 2 at 14:14
  • 5
    @Livy, the link speed is the link speed, and there is no slowing from 1 Gbps to 800 to 900 Gbps. If the link is 1 Gbps, then it works at that speed, or not at all. Your fallback is about copper 1 Gbps needing all four pairs, while 100 and 10 Mbps use two pairs, but the auto-negotiation is about what the end-device interfaces can do, not the cable quality. Fiber does not negotiate, but it will work for kilometers (depending on the fiber grade and transceivers used). – Ron Maupin May 2 at 14:42
  • 1
    From what I've heard, if a link cannot operate at 1000mbps, it will negotiate with the other end to fallback to 100mbps, and then 10mbps. You heard wrong (for Ethernet, at least). Autonegotation is about the end nodes' capabilities, not the cable. If the cable doesn't work it doesn't work. The only exception are Broadcom chips that can detect a two-pair link incapable of gigabit and renegotiate for 100 Mbit/s ("Ethernet @Wirespeed"). The negotiated link speed is a guaranteed constant value for the link. If the cable isn't up to it, it'll drop frames or fail altogether - no fallback ever. – Zac67 May 5 at 13:15
  • 2
    @Livy, you are trying to compare apples and zebras. Wired is different than wireless, and ethernet is a different protocol than Wi-Fi. – Ron Maupin May 5 at 13:17

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.