1

I work in research university, and we are moving to a new building. I suspect that new building will be equipped with mostly 1GbE copper, 10GbE copper, and possibly 10GbE-compatible fiber.

Is it reasonable to ask for 40GbE-compatible fiber for future-proofing?

My argument would be:

  1. We currently need 10GbE and they will put fiber in
  2. "Faster" fiber, i.e. OM4 MMF instead of OM3 MMF, is 20-30% more expensive
  3. But the labour cost is the major part of networking installation
  4. In the future we can install 40GbE switch and transceivers for small faster research VLAN linking few computers

3 Answers 3

1

Ask about OM4, but not for 40 Gb, for long runs of 25, 50, or 100. Price per Gb is best on 25 and 50, due to fewer lanes.

Although, fiber versus copper is a bigger difference. 25 Gb interfaces in some workstations or servers could be the right amount of bandwidth. Relatively inexpensive, until the copper to the switch closet needs to be replaced.

1
  • "Relatively inexpensive, until the copper [...] needs to be replaced" -- that is exactly what i am trying to communicate. We will outgrow (experience tells ~5y) 10GbE. Jan 30, 2020 at 14:44
0

Certainly individual desktops/laptops are never going to see 40 GbE speeds, but there are servers. I would say you can justify routing it to data center(s) and maybe some other lightweight server "closets", but not throughout. If you are linking servers within a data center, I would cross that bridge when you need it. At this point, I would skip the copper as much as possible and just go with fiber, even to desktops. You can lower your costs that way and offset the higher speed fiber where you need it.

2
  • "Certainly individual desktops/laptops are never going to see 40 GbE speeds, but there are servers." I remember that exact same argument about 1 Gbps ethernet: "It will never be used at the desktop, but servers can use it." That was back when the server bus topped out a 400 Mbps, and even servers could not fully use 1 Gbps ethernet. We now have 25GBase-T and 40GBase-T that can be run to workstations on Category-8 cabling (albeit a much shorter maximum channel length at 30 m than UTP at 100 m).
    – Ron Maupin
    Jan 29, 2020 at 23:54
  • 10GbE, among other things, allows us streaming data from storage into RAM and back. Even if we don't have enough RAIDed solid-state storage on the workstation. I have worked in building (built ~2008) where original plan called for 100Mbps copper in the walls Jan 30, 2020 at 14:41
0

Is it reasonable to ask for 40GbE-compatible fiber for future-proofing?

Yes. You definitely want to deploy the best available fiber to allow easy upgrades in the future. You should deploy OM4, OM5 or OS2 (for longer runs). Make also sure that there are enough individual strands for multi-lane PHYs, even if they are not terminated. The cost difference between different fiber grades is pretty much negligible when compared to the cost of deploying them. As John has pointed out, using PHYs with lower lane counts can also be much cheaper (single-lane 50G-BASE-SR vs. four-lane 40GBASE-SR4).

Terminating the fiber is a different thing. There likely is a large cost difference between terminating e.g. 10GBASE-SR and 40GBASE-SR4. However, which the proper fiber grade deployed you can upgrade the termination any time.

There can also be a large difference between 10GBASE-SR over MMF and 10GBASE-LR over SMF when using original SFP modules. With (much cheaper) compatible 3rd party modules, the SMF L-PHYs aren't significantly more expensive than the possibly multi-lane MMF S-PHYs.

Rule of thumb: if you want to or need to stick with original vendor SFPs, go for MMF as long as reasonably possible. If you can use 3rd party SFPs, seriously consider SMF for any deployed cable or even throughout.

Realistically planning ahead can save you a large deal of money later on without spending too much more today.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .