Design consideration – should I use a single fiber between two server rooms or split it into segments for easier maintenance/diagnostics/replacement? But splitting the fiber into multiple segments connected via adapters can significantly decrease the quality/speed, no?

.        enter image description here

Segment A is no problem. Segments B (fiber in ground) and C (fiber in building walls and cable channels) have problematic access and maintenance, each on its own.

Details: SFP to SFP, LC-LC, singlemode, DOM, WDM, simplex (due to space restrictions in section B). Planned speed is 1G. I must use purchased components, vendors here are currently not willing to arrive and adjust a cable, they are busy with large-volume orders.

  • 1
    This kind of borders on personal preference (which is off limits here) but I would never use adapters or couplers to connect 3 shorter cables together. I'd find a way to run a single cable securely (like using some kind of conduit to house the cable until it emerges at each end). Fewer single points of failure, etc. this way. – Jesse P. Oct 3 '20 at 21:01
  • @JesseP. – I was also thinking about a compromise – two cables: AB + C. Still one coupler. Maybe the better question should be "could couplers cause significant harm?" How would a pro design this? (currently without possibility of separate conduit in tight section B) – miroxlav Oct 3 '20 at 21:06
  • @JesseP. "I'd find a way to run a single cable securely (like using some kind of conduit to house the cable until it emerges at each end)." That may not be legal. Outside cable is restricted to no more than 50 feet inside a building, and extending it more can get the building red-tagged and shut down. Splicing cable should be no problem, as long as it is matching cable, just calculate the loss budget correctly. – Ron Maupin Oct 3 '20 at 21:47
  • @JesseP., segment C violates the length of outside cable that can be run inside, so it would need to be coupled or a splice, not a continuous cable run. Building do get red-tagged preventing occupation until a problem is resolved. I have seen that multiple times. The risk of fines and losing a lot of money while the building cannot be used should be enough to prevent such things, but people often ignore it. – Ron Maupin Oct 3 '20 at 22:32
  • @RonMaupin True. I guess terminating the B cable to fiber trays on each end and then running patch cables from B to A and B to C is really the only option. – Jesse P. Oct 3 '20 at 22:36

should I use a single fiber between two server rooms or split it into segments for easier maintenance/replacement?

Fiber doesn't require any maintenance. If you provide decent protection (duct, tube) it'll last forever. Always terminate fiber in a panel and use patch cables to connect equipment - an exposed cable is a vulnerable cable.

Only if more flexibility is required it is reasonable to use more connection points than directly necessary. Each connection is a potential breaking point, so less is better.

Also, local regulations may require using different fiber types outside and inside buildings. So, you'd terminate on the edge (in a panel) and patch in between.

But splitting the fiber into multiple segments connected via adapters can significantly decrease the quality/speed, no?

Yes, each connection loses some power and causes some attenuation, depending on connector quality, type and cleanness. Additionally, the pigtail splices cause some attenuation as well.

However, a short fiber run's power budget easily copes with those losses - keep in mind that the reach is in the hundreds (for multi-mode) or thousands of meters (for single-mode). 90 m is nothing, even with several connectors in the path.

You've specified SMF and Gigabit, so you're most likely looking at 1000BASE-LX10. Its power budget per 802.3 is 8.0 dB. Subtracting .04 dB for 100 m OS2 fiber, six connections of .2 dB each (three patch cables), you've got a total loss of 1.24 dB - plenty of headroom.

  • Thank you. I realized only after you reacted that by mentioning "maintenance" I used wrong term. I actually meant diagnostics, i.e. possibility to disconnect a section in case of faulty connection and check it apart of others to find which one has a fault. – miroxlav Oct 4 '20 at 11:30
  • Pretty much the same is true for diagnostics: if the cable isn't exposed or stressed, there's no need to diagnose it. – Zac67 Oct 4 '20 at 12:27
  • I get your point. In my case I am afraid that segments B and C can be occasionally exposed to work of 3rd parties on infrastructure laid there and in this setting I cannot satisfactorily prevent a chance of stressful manipulation with the cable. – miroxlav Oct 4 '20 at 12:34
  • Then, by all means protect the fiber by extra tubing or similar. Using an extra termination point is a trade-off between additional flexibility (also with fixing/circumventing problems) and introducing additional points of failure. Also, with 3rd party access to a potential termination, an exposed connector might be an extra security risk. – Zac67 Oct 4 '20 at 12:59

Every connection introduces signal loss, albeit small. (at the distance you're talking, you could string together patch cables and still maintain enough signal for SR signaling.)

It's common practice to have some sort of "landing" at the edge of each building -- a telco room, patch box in a plenum, etc. That allows for termination of armored, direct burial cable, and a clean transition to inside wiring (i.e. plenum / riser cable.) The inside wiring for fiber is usually protected in conduit. Being in a conduit also makes it easier to pull, and eventually replace.

As for there only being room for one fiber, fiber optic cable -- the actual stand of glass -- is very tiny. What you think is only enough room for a single cable could probably support 12+ stands. (single mode, of course. never waste time burying/fishing MM, even at only ~90m)

  • Thanks for insights provided in the answer, they greatly complement the other answer. – miroxlav Oct 4 '20 at 12:19
  • The difference in diameter between 12-strand or 24-strand trunk cable intended for MPO/MTP termination, and single-strand cable is barely perceivable to a human. You have to actually measure it. A 24-strand trunk cable is much smaller than a typical pair of single-strand cables as used for LC/SC/MTRJ termination since each of the two fibers has its own core, cladding, buffer, and sleeve. If you thought 12-strand cable was 6 times the size of a typical dual-fibre cable, you are wrong. You would probably be able to run 8 strands, which keeps your options open up to at least 800 Gbit/s. – Jörg W Mittag Oct 4 '20 at 13:18
  • (… and even more with WDM). – Jörg W Mittag Oct 4 '20 at 13:20

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.