Patrick's 2021 Fiber Build-out
- Thread starter Patrick
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Not yet, at least not really. Broadcom has designed some "Optical Power Converters" that will supply the receiving side with with a few hundred miliwatts
Optical Power Converters (broadcom.com)
The IEEE has worked on this as well:
150-W Power-Over-Fiber Using Double-Clad Fibers | IEEE Journals & Magazine | IEEE Xplore
But, really, it's a fairly inefficient way to transmit power
Essentially, 80% of the power is lost
Last edited:
Well…full fiber for home (IMO) isn’t really needed unless you have multiple rooms with 40Gbit+ equipment, and you are really planning to shoot that much data down the pipes between them as part of your setup - which is quite a niche case. For most homes (from small apartment to the McMansions) fiber is more of a "good idea" to, get reliable networking where signal from multiple Wifi APs can't traverse through (masonry walls, heavy EM/Spectrum pollution), your conduits have limited clearance, and you have so much noise in your electrical circuit that broadband over power line (HomePlug AV/AV2) is just not practical. Its more like a future-resistant backhaul between distribution points in a home network, using a media that is easier to thread through walls and conduits (or easier to hide under mouldings in walls). Depending on your home you might have to armor up the fiber (New York City rats will chew through OS2 like nobody's business) or hide it well.
Fiber of course comes with its own challenges. Is it LC? Is it SC? Is it multimode (MM)? Is it Singlemode (SM)? Simplex or duplex? Which fiber grade is it? Whats the minimum diameter for running them around corners? Is it armored? What's your conduit path going to look like and how do you plan to mount it? Is the space "yours" and what's your remediation if the mounting messes up the look on the surface? Do you want to take up the challenge of terminating them yourself (which is not easy and not cheap), or do you buy them pre-measured? Personally I went with singlemode OS2. The pricing versus OM4 is similar, the smaller diameter means tighter minimum curvative so it round corners better, and frankly, OS2 can easily pass 40Gbit+ on a single duplex LC pair, so it's future resistant.
Then you have to think about the fiber/copper interfacing (media converters and the like - do remember that a 10GbE SFP+ to copper media converter runs around 600 USD, and you might need pairs of devices), and the requirements for PoE - Power over Ethernet. If you have surveillance equipment or telecom equipment that runs off POE, at some point you'll need to integrate a POE switch (or power injector) and run copper through, and a power circuit to draw power from. If you have switches at each endpoint, will they generate its own noise/power/heat issues, and how does it play with existing cabling? I've been to newly built/green-field homes (made in the past 10 years) where one of the big selling points is that it's pre-bundled with existing CAT6E, and all of the RJ45 in the house terminates in what is essentially a 10MBit hub in a closet...with no ventilation, power circuit, proximity to the telco demarc) or accommodations for a 19" rack. They build homes but they sure can’t give you a network. In this case you might have to pull a fiber from your demarc, re-terminate/pop a gig switch into that closet, put in a rack, install some foam noise deadeners and wire a power outlet in there (preferably one that can power a giant low RPM fan to get some airflow going). If the conduit allows it you might be able to borrow the wire run for the CAT6e and use it to pull a fiber pair through.
Most consumer equipment do not/will not come with SFP+ cages (although I would be surprised and delighted if my next Wifi6 router comes with them), and even if the optic transceivers are built-in and it's at a reasonable cost (look at the Allied telesis AT29M2-SC that's available for 10 USD meant for HP 3/4 series thin clients and able to be shoehorned in certain M.2 Key A+E equipped devices), you might still have to do an SC/LC, MM/SM or fiber/copper media transition. AFAIK the NIC on that card (good old Broadcom Tigon) is considered obsolete and not supported on VMWare ESXi 7, which is not really a full solution. Even when some hardware comes with its own 10Gbit NIC built-in, very often the hardware integrator did not even bother to wire it up (a good example would be the 10Gbit NIC built into each and every AMD Ryzen embedded chip - which is almost never connected to a PHY and then a switchport - not exactly a vote of confidence on AMD's relatively new embedded NIC).
Things to definitely think about when you fiber up a home, or a small business campus.
Last edited:
I believe that gaming = cabling. More reliable.
Aren't Routers/Switches technically a "fiber/copper interface"? For PoE, I thought that you could get fiber to some PoE specialized Switch with a SPF Port for uplink like the MikroTik RB260GSP and call it a day.
I know consumer stuff has no SFP/SFP+, but theorically they COULD, and there has been a few Server/Workstation Motherboards that came with those like several Supermicro (Mostly Xeon D, but also a few on the Xeon E5 / Scalable platform) or the Gigabyte GA-6PXSVT. Rare, but it has been done.
And I was a big critic of AMD not exposing Zen builtin 10G MACs on any other line but EPYC/Ryzen Embedded since that being part of Zen becoming public. I was even more dissapointed about Motherboard vendors NOT exposing those even when available, and I recall Patrick mentioning something around the lines of most vendors having a full network stack that included the Hardware NICs, so they didn't want to replace them with AMD new one. But, ironically, on consumer, where a dull 10G NIC without SR-IOV or offloading capabilities of any kind could be reasonable, AMD decided to not expose it at all, which is where it would actually see more use due to not being part of a full stack, nor behind on features.
AMD had a chance to Conroe the consumer networking market besides CPUs. Instead, we got some lame middle-speed 2.5G and 5G on dedicated PCIe NIC ICs instead of using AMD builtin stuff... Still salty about that.
Possibly - but for gaming if your wifi was acting up you would've usually dealt with it ASAP - even if it meant dragging copper ethernet across rooms with couplers. Of course, instead of running ethernet you'll just mount some OS2 up on the conduit (or gluepad it to the ceiling mouldings), put a media converter on both sides and go to town as-is.
As for whether its a fiber copper interface? It depends on what gear that you have but if it's for a home situation, chances are, you are not going to have a full managed switch in each room (each needing power, cooling and emitting noise) with a couple of SFP+ cages on the chassis. It might be just an 8 port passively cooled 30 dollar "Microcenter special" unmanaged switch that only consume copper GigE at your endpoints...or maybe the room already have CAT6E/7 in place, and all you are doing is wiring between network closets in your house - save the copper for the "last mile" and use fiber for backhaul between floor(s). In which case, your fiber copper interface might be the network equipment in your house, or it could simply be a media converter that sits in front of a particular switchport in your older/less capable network switch.
Yeah, you can totally do SFP/SFP+ - hell, you can buy an M2 SFP (1GBit) adapter that would work with the Wyse 5070 thin client (and other small machines), but it's also 249 dollars, double what a 5070 is worth on eBay. Unless you have economies of scale and large demand, it'll be expensive. And frankly, unless it's for 10Gbit (SFP+) or above, it's pointless, and 2.5/5/10 are on copper these days (yeah, I know, 2.5/5 are rather pointless). Pluggable optics are probably limited in its lifespan - as soon as Intel and other vendors integrate photonics directly onto the silicon it'll be on its way out...figure 2024 or so. I kinda doubt anyone will push hard for SFP+ between now and then.
As for AMD and that entire 10GbE non-adoption? Supposedy that 10GbE IP block was not originally from AMD, and whomever they licensed it from wants a royalty check for each NIC being used, so AMD simply charge the integrators more (which covers the licensing costs) if they want to use it. it's a fairly simple NIC without the fancy features that all the kids are up in arms about (like SRIOV), and it’s no better than the Intel NICs from 7 years ago (like the i82598, which is cheap with mature drivers at this point). Oddly enough none of them wants to do so, and they simply went to Realtek or Intel for gigabit NICs instead (the fact that the Realtek 8111 supports their DASH VPro alternative versus their own…which doesn’t…didn’t exactly help their own cause). AFAIK that 10GbE NIC on the Raven Ridge chip is just deadweight (you know something is odd when both Synology and Qnap did not expose the 10GbE ports on their Ryzen V1500B based NASes). Yeah, I know, pity. Let's hope that Xilinx's SolarFlare division can toss AMD a good 10Gbit NIC to integrate into their future APUs. For AMD’s sake this should be done, like, yesterday if they want to stay ahead of Intel.
Last edited:
Actually, I am referring to those of us who doesn’t need to turn our homes into a hardcore data center, and frankly, he’s not wiring a home - it’s a business campus…so not quite the same.
I also didn’t say not to wire up for fiber - I said that there are issues to consider before you pull the trigger, and strategies to use while wiring it up. If you have existing conduits to run copper, if you have CAT6E already in the house, if you don’t need metric truckloads of bandwidth between rooms, if you have POE gear all over the place which is otherwise already centrally powered from a single location (which might require running a parallel fiber network, or you need injectors or sub-switches) then there’s not much urgency to run fiber.
When I wired my apartment up for fiber I re-did the backhaul from the demarc (where the ISP equipment is located and where the wireless router(s) are located) to the area where I kept my servers (the VM host and the NAS/iSCSI target), which was meant to be quiet and unnoticable. I needed media converters on both sides (since the equipment on both sides are still copper/Gigabit ethernet), which combined with the fiber itself meant that it was overall much more expensive (on a per-meter basis) than the copper ethernet that I could’ve run.
I ran fiber simply because it looked much better/was easier to hide, I can put surge protectors/UPSes in the living room circuit to protect my gear, and it was done to replace a Powerline AV 1200 setup that was both unreliable and slow (old circuits and poorly isolated fusebox). The hope is that 10Gbit switches/media converters will eventually drop in price and my setup will go to 10 or even 40/100 one day (40 or 100GBase-LR4 on both ends of that OS2? eeeyeah, not for a while and not at 400+ dollars per transceiver) - I am not sure if it’s really needed right now, but it’s nice to have some built-in margin just in case. The connectivity between my t740 and the HP MSG7 N40L is already 40GbE.
Besides, @Patrick already has enough sunk cost in this project to not back out of it now.
Last edited:
I think for most people in the past year and a half there were some crossovers
Ehhhh…to the point where you need to run fiber in your house to deal with bandwidth demand…?I think for most people in the past year and a half there were some crossovers
I mean...I don't really need 2 dual-socket servers, a disk shelf, 6 switches, 4 access points, god knows how many servers and computers in my lab, and at least 1 desktop computer on every floor in my house.. but I have it
PS Don't tell my fiance
The only real reason I don't have fiber strung around the place is because I am busy and lazy.
So a bit of an update. The long 144 fiber single mode spool seems like UPS lost even on an in-state shipment.
Ordered a new longer spool that was 110ft longer because it was the fastest one I could get given current lead times. Apparently, that one because it is ~25% longer needs a bigger and more robust spool. Between the cable length and bigger spool, it weighs ~145lbs without a box. As a result, it will go over the standard FedEx / UPS ground shipping weight. So now it is heading here freight.
Meanwhile, the copper network is pretty well sorted at this point. We re-keyed all of the phone jacks that were Cat5e in the wall to Ethernet and RJ45. Testing those out, all are doing 5Gbase-T speeds. Some are doing 10Gbase-T. Interestingly enough the 10Gbase-T ones even at ~75ft tend to be the ones run using outdoor cable through the walls. Odd, but can only complain so much. We are going to run a bunch more copper for cameras/ APs but at least some baseline capability now with what is here.
Hopefully tomorrow we start actually pulling cable and the new huge spool arrives Tuesday.
Ordered a new longer spool that was 110ft longer because it was the fastest one I could get given current lead times. Apparently, that one because it is ~25% longer needs a bigger and more robust spool. Between the cable length and bigger spool, it weighs ~145lbs without a box. As a result, it will go over the standard FedEx / UPS ground shipping weight. So now it is heading here freight.
Meanwhile, the copper network is pretty well sorted at this point. We re-keyed all of the phone jacks that were Cat5e in the wall to Ethernet and RJ45. Testing those out, all are doing 5Gbase-T speeds. Some are doing 10Gbase-T. Interestingly enough the 10Gbase-T ones even at ~75ft tend to be the ones run using outdoor cable through the walls. Odd, but can only complain so much. We are going to run a bunch more copper for cameras/ APs but at least some baseline capability now with what is here.
Hopefully tomorrow we start actually pulling cable and the new huge spool arrives Tuesday.
Patrick's project is a good motivator. Instead of relying on my switch telling me the lengths of my longest drops, I confirmed through their sheathing today. TDR calculated 3 m more. Purchased housing and cassettes a couple of days ago.
On today's STH video : man builds data center meet-me room, pulls a John Cusack with a Corning toaster oven, and worries about getting enough fiber.
It is happening!
Hi @Patrick
I'm confused a little by this. I've never seen fiber pulled through the wall like that bare.
Even with pre-termed MPO/MTP or even LC connectors, I've only ever seen fiber run through a building inside of innerduct:
Sometimes inside of conduit like this (hence "inner"duct)
Sometimes totally bare like this:
and just run through the plenum along side the copper plant's J hooks....or even just thrown on the ceiling tiles...
Heck, I've even seen people run fiber through split-loom they bought at home depot
Corning even has a whole whitepaper on how to do it properly:
005-011.pdf (corning.com)
However, I've never seen it run bare. In any case, from the perspective of future expansion and physical security of the cabling, why did you choose to run the cable the way you did?
EDIT:
I AM REALLY STUPID IS THAT ARMORED CABLING??
Last edited:
Somewhat the idea of this is to not pull again.
The cable is not full armored cable like you have there, but there is the outer sheath of the cable, then reinforcement, then the outer MTP cable covers, then what is inside of those. By the time we have passed the split-loom-like pulling section, it is then suspended in the pulls between the ceiling and roof with pulleys for most of its journey. We are then strapping it in. Only one of the runs is >35M and usually, we are only dealing with 1-2m to air spaces on either end so the cable is not even getting scratched.
Conduit would be better if we wanted to pull more, but the studio will have 288 fibers. The offices will each have 240 fibers and so forth so I do not really foresee a need in the next 10-15 years to pull any additional cable. Part of pulling big bundles is to have redundancy and such in the event there is future damage to the cables or connectors.
The first part with the orange may look like normal duplex fiber, but it is actually a pulling assembly with a spinning end so that the fiber does not get twisted by the pulling on its journey.
You are correct, we could go do conduit armored cabling and more, but at some point, there is only so much budget to spend putting fiber in a house.
The cable is not full armored cable like you have there, but there is the outer sheath of the cable, then reinforcement, then the outer MTP cable covers, then what is inside of those. By the time we have passed the split-loom-like pulling section, it is then suspended in the pulls between the ceiling and roof with pulleys for most of its journey. We are then strapping it in. Only one of the runs is >35M and usually, we are only dealing with 1-2m to air spaces on either end so the cable is not even getting scratched.
Conduit would be better if we wanted to pull more, but the studio will have 288 fibers. The offices will each have 240 fibers and so forth so I do not really foresee a need in the next 10-15 years to pull any additional cable. Part of pulling big bundles is to have redundancy and such in the event there is future damage to the cables or connectors.
The first part with the orange may look like normal duplex fiber, but it is actually a pulling assembly with a spinning end so that the fiber does not get twisted by the pulling on its journey.
You are correct, we could go do conduit armored cabling and more, but at some point, there is only so much budget to spend putting fiber in a house.