If you get the chance, always run more extra network fiber cabling | Hacker News Reader

If you get the chance, always run more extra network fiber cabling

utcc.utoronto.ca

230 pointshggh244 comments

Yup, always run extra everything. At least you can splice fiber now. And be careful of the minimum bend radius.

Once, in the 90s, we were having intermittent network failures in our data center. I kept trying to troubleshoot it with the fluke, but the problem kept moving. When I pulled up the raised floor, I discovered that rats were eating the exterior sheath of the network cables. That was some fun troubleshooting!

_minimum_ bend?

minimum bend _radius_

A straight cable has an infinite radius, the more bend the smaller the radius

Though if it is on or under the surface of the earth, “straight” will be a bend radius of around 6,370km. We don’t make a lot of buildings that deal with this but transcontinental or transoceanic cables certainly do. If someone designed a fiber that required absolutely no bend in order to work you’d have to use it in buildings or dig much deeper holes.

There was an encoding mechanism proposed about 10-15 years ago that used spirally polarized light to carry more channels, but it required the surface of the fiber to be polished to a much higher degree than existing cables in order for the light to go around bends properly.

If you’re using the planet as your “flat surface” then sure. If, however, you’re willing to deal with exiting the atmosphere at each end, you can use Real Straightness. But I don’t know anyone running a single segment for that distance.

I’m sure there are some microwave antennas still out there doing the Lord’s Work. At least in the Plains states where hills are low and putting antennas on two of them gives you some extra distance. How far do microwaves bend over the horizon?

Don't they generally do the opposite of bend over the horizon? Two towers that are observed (visible wavelength = tiny Fresnel zone) to have line-of-site can easily be obstructed (microwave = huge Fresnel zone).

https://en.wikipedia.org/wiki/Fresnel_zone

> How far do microwaves bend over the horizon?

that depends on the weather and frequency among other things

https://en.wikipedia.org/wiki/Line-of-sight_propagation

If you're going out of the atmosphere may as well skip the fiber and just point lasers through the vacuum of space directly and reap the benefits of the faster speed of light through a vacuum vs glass!

works great until a group of space bats get in the way.

Minimum bend radius.

It’s usually 20x the outer diameter of the cable for fiber, a cable with a .3 inch OD has a 6 inch minimum bend radius.

What?

This is 100% wrong.

This is well standardized. As I posted elsewhere:

G.657A1/B1 = 10mm (half an inch)

G.657A2/B2 = 7.5mm (a little over a quarter inch)

G.657A3/B3 = 2.5mm (less than 1/8th of an inch)

The only cable you will find with a 6 inch bend radius is going to be armored interlocking fiber cable or something.

You can take 2mm G657.A3 fiber cable, wrap it around a pencil, and it will work 100% fine.

Even old G652 cable only had a 1 inch bend radius.

If I am wrong, then explain why this 24-strand plenum rated fiber optic cable has a minimum bend radius of 5.9”: https://leviton.com/content/dam/leviton/network-solutions/pr...

Part number: LTP12B024

I am running a project where the low voltage contractor is pulling this fiber cable into a conduit my electricians are installing.

I don’t doubt your bend radius claims on the fiber optic cables you listed, but there are far more types of fiber cable than the ones you listed.

There are not far more types of fiber cable, there are far more types of jackets :)

You are confusing the cable jacket with the fiber cable inside of it. This is the spec on the jacket they use and they didn't design this particular jacket to be flexible enough to survive flexing at the level the fiber cable inside of it can. Sure - not all jackets can be bent as much as the fiber. Some because they are not made to be as flexible (as here), some because it's basically impossible (armored interlocking jackets), etc.

I'm not sure what this changes?

The part number with suffix AB0403 (the SMF A1 fiber), if you cut the jacket, you should be able to bend the fiber cable inside at a radius of 10mm and have it be fine. It appears to be standard G.657A1 fiber cable inside.

There are really not far more types of fiber cable than i listed, and this spec sheet definitely does nothing to support that claim, since as i said, it has A1 fiber cable inside the cable jacket that will happily support a 10mm bend radius.

BTW, as an aside, I wouldn't use this stuff - just looking at this sheet and pricing at my normal fiber distributors - it is both overpriced and underspec'ed for its price. Which seems pretty typical for leviton :)

Most 24 standard I/O plenum cable is going to be 7.8mm (this is 9.9, so 30% thicker), and have less than half that bend radius even with non-flexible OFNP rated jackets. If you are being charged more than 2 bucks a foot for it, you should consider other options.

> There are not far more types of fiber cable, there are far more types of jackets :)

Ahhhh, thanks for clarifying! Now I understand what you’re saying, the jacket and fibers have different bend radiuses (radii). I appreciate you taking the time to explain :) also, noted about there only being a few types of fiber, with a multitude of jackets.

I’ll let my low voltage subcontractor know about the underwhelming Leviton cable, it’s a plan and spec public job so it’s possible the Leviton rep is friends with the engineer and got their cable flat specced, wouldn’t be the first time!

I’m only responsible for the raceway, it’s a 2500’ run between two college campuses separated by a highway that runs outside for a good chunk of the distance so it’s possible the jacket is beefy and inflexible for temp resistance. I’m much more familiar with electrical conductor jackets, THHN and XHHW-2, which are PVC and XLPE, respectively.

Fibers can be bent to a radius of an inch or so, that's the tolerated minimum bend radius.

I guess you could call it "the maximum sharpness tangency" or something like that, but that's not the standard verbiage.

G.657A1, which is basically the crappiest you would ever buy at this point, is not even half an inch bend radius.

Have fusion splicers come down in price? Can you recommend any?

You can have them second hand for ~500€ on ebay

Any advice on making sure you get a good splicer?

I know a little about fiber connectors, and the different connector modules for speed, but I am not really sure what I need for a splicer for fiber.

AliExpress sells some basic ones from $300-700 CAD ($200-500 USD) with reasonable reviews.

Whatever you do, don't try using the razor-style "hand splicers" and adhesive splice kits. Without a splicer that has a scope, you're just making bets each time that are difficult to test.

I've learned the hard way that it's just not worth it unless you can _see_ what you're working on.

For Multimode, they now make splices where you jam the two cleaved ends into the splice. It costs about $5 per splice.

You don't need them unless you are running a mile or two of fiber - mechanical fast connectors (IE no polish, etc) are less than 0.3db loss at this point.

Fusion splicing will still be 0.01db or better, but you will be well within the power budget of any transceiver you find with mechanical connectors, if we are talking home networking - even a 2km 10gb transceiver that is 20 bucks has a 6db power budget.

Heck, the 10km are cheaper than 2km at this point - 17bucks on amazon, power budget of 16db.

This is VERY true - especially for home use. You could have many db of loss and still have a perfectly functional 100g network.

I have a $1000 fusion splicer from Amazon, and it is fantastic. It is very straightforward to use, and for my home use it is perfect. If I did spicing as a job this wouldn't be the tool to use, but for occasional splices in my homelab it works great.

https://www.amazon.com/dp/B0CW2HGVT2?ref_=ppx_hzsearch_conn_...

Your typical Rodent Accelerated Transmission Signal And System Suppression

> Yup, always run extra everything

Cables ain't free, duct capacity is finite and duct rental from the local incumbent is costly too... Please calculate the financial optimum of pay now vs. pay later - taking into account growth, various forms of attrition, cost of capital, opportunity costs and appetite for risk. Or everyone would be running 1152 strands cables everywhere.

But then I see that from a telco perspective and, now that I've read the article, it seems to be from a small-scale hosting perspective - entirely different economics.

Or, take the money you would spend calculating that, and that's the amount you spend for your margin.

Haha, I'll tell that to the outside plant capacity planning team !

But seriously, I'm telco-biased, and it seems that the article has a small scale hosting... Different worlds. At small scale, sure - the bigger cable is a rounding error.

Duct capacity is finite, but fiber cables are tiny and they are very nearly free compared to the cost of having them run.

Not quite tiny - a basic armored (rats !) 12 strands cable may be around 10-15 millimeters, whereas similarly specced cables in the 500 strands range might reach 20-30 mm. In crowded Paris downtown, it adds up fast. On the countryside, that is a lot more weight to hang on poles.

    > Paris downtown

Where is downtown Paris? 1/2/3 arrondissement? I never heard the term used for Paris. Mostly, it is used to describe the Wall Street area of Manhattan (NYC) as the island is relatively north-south and Wall Street is on the southern end.

It’s also come to mean the densest part of any city. Even the town I group up in would call the densest part “downtown”, even though the only buildings taller than two stories were city hall and the historic theater.

Today I learned!

I always thought "downtown" and "uptown" were more like "south from X" and "north from X" where X is some reference point in context (e.g. "where I'm standing"). Turns out, it's more similar to "downwell" and "upwell", i.e. towards or away from some central point (gravity wells being spherical-ish and all).

Uptown can mean anything out of: geographically north, upstream on a river, higher elevation, or higher income and wealth of the residents.

It just so happens that for Manhattan which is the most famous uptown, all four of those indicators point in the same direction. In general I think an area would more likely be called an uptown if it met more of those criteria relative to some downtown. (And there would be correlation, wealthier residents would buy higher-elevation land.)

I notice that non US English-speaking tend to use the term CBD instead of downtown, my assumption is that it is because of the NYC thing you mention

British people would say city centre or town centre.

Australians are the ones who say CBD.

I've seen it in Canada, NZ and India as well. Def not AU specific.

New Orleans seems to like giving directions relative to it's CBD (just south of the French Quarter). It's otherwise not common here in the US.

In New Zealand we refer to it as the CBD, but you could also say city centre or town centre (or just “town”).

Not sure I have ever heard downtown being used to refer to part of a New Zealand city although I understand its meaning.

I don't think there's any association with New York. etymonline claims that "downtown" originally refers to the idea that the suburbs are at a higher elevation than the city proper. I don't know how much stock I'd put in that, but I trust it more than someone's random intuition about the geography of New York.

( https://www.etymonline.com/word/downtown : "The notion is originally literal, of suburbs built on heights around a city." )

It's the only available word for the concept in American English. I wouldn't recommend saying CBD; that would refer to cannabidiol or to marijuana in general.

Reading more about it, accounts vary. There are historians who posit it was in fact inspired by Manhattan, or maybe Boston, but it's not cut and dry.

Edit forgot link: https://downtownny.com/news/where-did-downtown-come-from/

I'm not especially impressed by the argument "if we assume it came from Manhattan, then we can know it came from Manhattan":

> Let’s suppose that Fosdick, incoherent from distress, forgot to include the words “to the” between “down” and “town” when describing the direction in which people were fleeing. With this in mind, we can assume that Lower Manhattan is in fact the original “downtown,” as posited by Schwarzer, and that the possible Bostonian origin of the word is simply a misunderstanding.

I'm also frustrated by the lack of sourcing here but the wiki article for downtown [1] also points to Manhattan, citing a book I don't have a copy of. Allegedly it was in use in Manhattan in the 1830s, which aligns with the date given by the etymology entry linked previously.

[1] https://en.m.wikipedia.org/wiki/Downtown

https://www.etymonline.com/word/downtown The word was in want of a closer look in fresh archives. On a closer search, it's older (at least 1780s), American, but not apparently a New York reference at first.

Interesting. I would use the ability to get 40x the strands for 4x the weight as evidence against extra strands adding up fast.

And while some people preach hard-wired everything I’d probably increasingly not bother at home. . I’ll have to see how much networking and audio stuff I even do given a kitchen fire with s

Maybe in an apartment. In a house, you want wired backhaul on access points, and wired streaming devices.

Why? Is this an actual problem? I have a bunch of wired connections (or at least had) but I'm not sure I ever saw a difference and I'm certainly not going to spend time or money retrofitting them for some Platonic ideal--ADDED, at least in recent years or outside of specific situations.

It's potentially a problem in a typical brick house in Europe.

I'm sure it's less of a problem in houses with other materials.

The answer is brick and concrete. You’d be amazed how well a load bearing wall in a high rise will block a signal.

Which are, of course, also difficult to retrofit with a lot of new wiring. Old plaster lath is something of an issue too but not nearly as much and is easier to retrofit--or just replace.

For me it's the metal studs in my walls. WiFi reception is truly awful. Fortunately it was prewired with Cat5e. But that will need to be replaced some day.

Latency. WiFi latency can spike which ruins video calls. Simple speed (given lots of time) on the other hand is rarely a problem.

There’s also the problem of sometimes, video calls just stutter and seize up and it’s the internet’s fault, not the local network. But with a wired connection, I am immediately confident the problem is not in my network. With WiFi, who knows, maybe it’s my fault. I used to continuously ping the router so I could verify this, and indeed ping spikes would happen. Now with a wired connection I don’t worry about this.

But yeah if you run no latency-sensitive applications, WiFi might be fine.

WiFi is convenient and (usually) cheaper (esp if it's a retrofit), but not deterministic. And less 'cluttered' I guess, if that's your pet peeve. All sorts of things can muck around with Wifi speed, reliability and latency and most people have little or no ability to troubleshoot it effectively.

I retrofit wired my house[1] about 20 years ago; I only have had a small handful of issues with it in that time (all basically mechanical fatigue of jacks). If I have speed, latency or connectivity issues, I can be reasonably sure the last thing I have to debug is the physical network.

[1] Copper, not fiber, but no runs are over 50m. Works to 10G so far. I used a Leviton residential structured wiring system (network, coax and phone at the time; the last 2 long since obsoleted). I did a lot myself since I'd done enough quality time in datacenters to be adequately proficient with low voltage Cat5. Not cheap but not extravagant.

I just had luck and have my ISP modem close to my desktop PC and desk so I run cable to desktop, laptop dock, TV and consoles all on the same wall.

So worth it as in apartment everyone and everything is blasting wifi. I see like 20+ SSIDs.

"And be careful of the minimum bend radius."

The minimum bend radius of any good fiber is way less than cat6 at this point.

G.657A1/B1 = 10mm

G.657A2/B2 = 7.5mm

G.657A3/B3= 2.5mm

CAT6 = 25mm

Flexible cat6 = 15mm.

Fiber surpassed copper for flexibility almost 2 decades ago (G.657 was introduced in 2006, but let's call it 5 years before all forms became common).

Even G.652 was only 25mm and standardized in 1984.

It takes less force to bend, though.

could be solved by running reinforce wire along it

> could be solved by running reinforce wire along it

You could, but you might as well just use armoured fibre if you're doing that !

Some of the low core-count CSTA stuff out there these days could almost have you thinking its unarmored if you didn't know. Not at all bulky.

for an example of how good modern g.657 bend loss insensitive fiber is, for ftth applications and drop/last hundred meters, here's an example of it being abused with a staple gun and the resulting loss in dB.

https://www.youtube.com/watch?v=h4wzJSInvww

the above is from 12 years ago

I've seen some g.657.b3 patch cords and work done for FTTH installs that are absolutely atrocious and yet continue to function properly.

This is proprietary technology and is definitely NOT representative of normal fibre cables. I know because I had had to deal with contractors doing exactly this, which crushed the cables we ran. There are many things which could be better if we were willing to pay higher prices for them. At the end of the day, cheap and good enough usually wins.

This is just G.657B3.

They are all proprietary in their coatings for sure, but it's made to a standard. The only thing is that it's not guaranteed compatible with old G.652 fiber if you try to fusion splice it.

You can get G.657A3 that is, if you want.

But this is representative of A3/B3 cables for sure, and easy to get.

AT&T fiber and Google Fiber actually use this particular stuff in their standard deployments.

Under the covers, most of the glass is usually corning clearcurve zbl (OFS actually makes their own i believe, but for others, less common).

The most common consumer cable these days is G.657A2/B2, which is easily capable of what you see in the video.

As for pricing, i'm not sure what you are paying, but this particular stuff is also easy to get (for a consumer) at 20 cents a foot or less. 1000ft is usually 200 bucks or better. Their are idiots on ebay trying to sell leftovers from fiber installs, but if you go to any really fiber cable seller, that's what you'll pay.

Here are totally custom cuts of 1.6mm outer for 0.25cents/foot: https://www.fiberinstrumentsales.com/tlc-simplex-sm-clearcur...

if you want 3mm outer, it's 0.35cents/foot: https://www.fiberinstrumentsales.com/tlc-simplex-sm-clearcur...

FIS is usually expensive, too. I didn't look hard since that seemed "cheap enough" already.

You can certainly get A1/B1 cable for like 18 cents/foot instead of 20, but if your contractors are crushing cables, seems a bit penny wise, pound foolish.

I also wouldn't run simplex cable, as it doesn't make sense anymore.

As expected, 6 fiber versions of the above cables are only marginally more expensive.

To be real frank if you're using FIS as a benchmark of price and market availability of product, you have already lost. Everything on there is 300% marked up from the price they buy it from the vendors in China and Taiwan. FIS is for enterprise end-user customers that don't know any better, not service providers who buy fiber stuff in quantity.

g.657.b3 is an industry standard and not proprietary, and after close to 14 years being on the market, is the default for ftth drop cables and inside patch cables now in the last mile.

This video is a cool demo. Trying to eyeball it, it seems all of the bend radii (except one) are over 10-20mm so this demo isn't that impressive. The only exception is the very last bend he makes with the weight where the radius is probably around 5mm. I find this ironic that when making this last bend is exactly when the video abruptly ends with some video corruption artifacts. As if they streamed the recording over this fiber :-)

As i'm sure you know - for invisible installs, they don't even bother to have an outer cable anymore, just the 250um fiber with no sheath running bare on the walls.

Kinda insane.

I accidentally put a knot in a fiber, and it still worked. I'm still using it today, 3 years later (after untying the knot, ofc).

And clean the fiber connector ends with a special cloth or cleaning liquid. Once, I had an intermittent link due to dust particles.

Fun fact 1: When you don't have a proper OTDR fiber tester to check for breaks, you can shine a laser pointer down the fiber and see it projected out on a sheet of paper on the other end (you might have to call the person on the other end).

Fun fact 2: If the other end is plugged into a transceiver, don't look at the fiber end directly; instead, look at it through your phone camera. If all is working correctly, you should be able to see a faint purple light out of one of the strands.

P.S. always remember to roll it if no link initially :)

The last one may not work anymore. A ton of phones nowadays have IR filters.

They all do and always have, they just aren't perfect.

Really? I thought they usually used to have no filters, especially on the front camera. Well of course the lenses always filtered some IR light, but I wouldn't call that a filter, they also "filter" non IR light. A filter to me is a dedicated element or coating that prevents the majority of IR to be transmitted.

works on my iphone 13

> If the other end is plugged into a transceiver, don't look at the fiber end directly

Let me fix that for you.

Don't look at the fiber end directly, always assume it is plugged in unless you can see the other end yourself with your own eyes.

Fibre light meters can be bought cheaply for a basic troubleshooting bag kit-out if you don't fancy splashing out on a fancy device.

Yeah, we got a new house built 4 years ago and one of the biggest regrets is not enough ethernet outlets (running fibre in the house isn't practical yet) and power outlets in various places. It is really hard to think of all the places you would want them ahead of time unfortunately and as each has a cost you don't want to incur it unnecessarily.

But my biggest regrets were:

- Only a single ethernet port in the basement. Then the kid wanted a gaming station and we moved where the TV was. I should have put like 4 down there.

- No ethernet ports in the garage, I should have put in one for an AP.

- 4 ceiling APs instead of just two in the main part of the house. I over-estimated how much coverage I would get from ceiling APs and thus I have some APs hidden under furniture to ensure 100% house coverage.

- Lack of multiple circuits in the garage, even better a separate sub-panel with 6 outlets. I took up wood working and with a single circuit and 2 outlets was insufficient. That cost me $1200 for the sub-panel.

- Multiple outlets on the back and side of the house - I would have done two at the front on each side and two at the back on side each and one on each side of the house. I have a single outlet at the front and back and that is just not sufficient for lights, decorations and patio devices.

Wow, we recently built a house and this reads just like my regret list!

I think because there are less code specifications for the garage, builders just don't bother.

> Only a single ethernet port in the basement. Then the kid wanted a gaming station and we moved where the TV was. I should have put like 4 down there.

Is it so bad to put in a switch?

It does sort of suck because I have cabling running along the baseboards. Also I have 10G networking so the switches are costly and hot at least this generation.

How would multiple runs have helped you with that problem? Either they’re all in the same spot, and a switch would solve it, or they’re on all different walls and you’d still likely have to run cables to the things.

That's why pretty much every office space/building over here has cable trunking/channels on the walls instead of fixed wiring, eg:

https://www.elba.si/izdelki/parapetni-kanali/

https://www.obo.si/izdelki/izdelkiinstalacije-v-zgradbah/bis...

Need one more outlet? Remove the plastic cover, insert cable, insert outlet, cut the plastic cover to make space for outlet, and you're done.

That would be fine in the garage, but few people would want this anywhere else in their home.

There's a wood effect version, but I suspect it would make a home look like a 1980s office: https://homemaster.techinfus.com/en/kabel-kanaly-vidy-i-razm...

In the us wiremold is the equivalent product.

> Only a single ethernet port in the basement. Then the kid wanted a gaming station and we moved where the TV was. I should have put like 4 down there.

Just drop a switch down there.

I have, but now I have cable running along the base boards. That isn't optimal. Also I have 10G networking, so the ethernet switches are hot and expensive.

You don't need an expensive switch. Five port 2.5G + SFP+ with poe can be pretty affordable and shouldn't run very hot.

That downgrades transfer speeds though to just 2.5G.

Which is fine? It's getting capped to 3 at your modem anyways... And the difference here is reaching the "doesn't matter" phase.

Like - 50GB at 3gbps is ~130 seconds, ~160 seconds at 2.5gbps, and ~400 seconds at 1gbps

You're always waiting at least a couple minutes, and never more than 10 for that download.

Personally - I have a couple 10gbps switches that run my backbone in my house (NAS, k8s cluster, HomeAssistant, etc) and I have a 2gbps symmetric wan, but all of the offices (my wife and I both work from home, plus the media room, plus the kid's playroom) all just run 1gbps switches.

They're cheap, normal office work basically never saturates those speeds (even multiple conference calls with video is a-ok), and there's just not enough value in bumping up to 10gbps across the whole house compared to the equipment cost.

The 10gbps switches live in the basement and just connect those zones.

You are completely right that I can always just wait. I prefer not to though... there are a lot of times when it is amazing to have 10G between my main machines.

The devices attached to the 2.5G switch, will be limited to 2.5G per connected port. There won't be any slow down on any other connections. If you still want 10G in the basement, put in a 10G SFP+ switch (they can be had for $15-20 per port and use DACs to control the heat some). You'll only have one shared 10Gbps to the rest of the house, but that is really unlikely to be saturated.

I was thinking about this and it is very expensive. And do I need 10G really? I do not think so. I did calculated few things when I was thinking about 10G and most probably I will not saturate 1G for few years. And with phones ans laptops mostly WiFi is used anyways.

So can I ask you what is the advantage for using 10G equipment in your case? Even something really dull like network card costs around 100€.

10G is awesome for (1) downloading steam games - many are +50GB these days, (2) downloading Ollama models, (3) downloading ISOs, (4) downloading app/OS updates and (5) I have a NAS also on the 10G network for media backups.

I have symmetric 3Gbps fibre to the home via Bell Canada and it reports 4.5Gbps via fast.com.

If I get this new job I'm definitely upgrading everything to 10G.

My PCIe5 SSD can do 14 GB/s, or about 112 gbit/sec.

Did some research. 10gbe cards are about $30, 100gbe qsfp28 cards are about $300, 100gbit 100m transciever about $100.

I’ve got Virgin cable and can pull 6Gb a minute from fast.com. Your hypothetical 50Gb steam game would take a whole 9 minutes assuming steam can saturate my connection. So that’s like enough time to hit start download , make a brew and come back to the pc. Why bother.

    > Also I have 10G networking, so the ethernet switches are hot and expensive.

I never understand why anyone needs anything faster than 1G (other than to flex for Internet points). Seriously, how often do you need to move data faster than 100MB/s? Almost never. FYI: A 2.5G switch should be fine. There are many fan-less options here: https://www.servethehome.com/the-ultimate-cheap-2-5gbe-switc...

DLT drives can saturate a 1Gbps link. Makes it hard to prevent shoe shining when backing up file servers.

If you have a NAS, for one. Fill it with SSDs and you can even saturate 10Gbps.

"I never understand why anyone needs anything faster than 1G".

i just upgraded my home network from 40g to 100g. g, not m. Anyone who does video knows the pain of 10g.

"Anyone" is of course a shorthand for "nearly everyone, but of course there are some exceptional people with exceptional needs".

I don't think your situation is common.

XikeStor SKS8300-8X costs about 100$, can do L3 routing, has 8 10G SFP+ ports, no fan and consumes ~9W. I'm running it in my homelab without any issues.

Friendly reminder: newer modem, router (especially for wireless), cat cable and fast speed switches really make a difference. I had the fastest speed my ISP could provide, but then discovered I needed to upgrade some of my equipment. It made all the difference in the world, especially cat 6e cable.

> Only a single ethernet port in the basement. Then the kid wanted a gaming station and we moved where the TV was. I should have put like 4 down there.

https://store.ui.com/us/en/category/all-switching/products/u...

Unless you meant with regards to routing the cables throughout the space, like now it’s on the opposite wall from the port.

I’ve definitely shoved cables under baseboards, but you can also pull them off and the run the cable in the wall behind them. You don’t even need to patch the drywall then. You can also get baseboards (that look nice in a home) then double as cable runs.

> You can also get baseboards (that look nice in a home) then double as cable runs.

Interesting. That would be great in the basement. I will look into it.

Dont forget Powerline adapters - if you have it all on one loop they are surprisingly effective if you have extra power sockets!

If you go that route MOCA over your cable tv coax works very well too!

I recently bought my first house, and there was only one coax outlet in the entire place, and zero ethernet. I wanted to move where the coax terminated, and after some searching around, found MOCA. Had never heard of it before, but a hundred bucks or so later, and it's a fantastic solution.

for those of you using MOCA, what kind of bandwidth do you get? And what kit would you recommend?

I ask, as my brother-in-law just finished a house where he had asked the contractor to put in internet cabling everywhere, and... drum roll... yes the contractor put in coax everywhere. I tried powerline adapters, but could only get ~30-40mbps (he has a 1gbps fiber connection from the ISP) over them.

I use the MoCA 2.5 adapter by goCoax. I won't say I recommend it since I haven't compared it to other adapters, but it appears to do what I want it to do and I get about 2Gbps through the adapters. Every one I could find seemed to be around the same price point ($60-$80 per adapter) so I just went with a well reviewed one.

I would definitely recommend if fishing cat6 is not an option and you would need to open up walls. Unless your BIL has a lot of kids and their friends over who all like to stream different 4k movies from a NAS, I can't imagine any normal user with 1gbps WAN finding 2gbps LAN insufficient.

I think powerline adapters might work very well in the UK, where sockets on each floor are connected in a ring.

In my small apartment in Denmark, with new wiring, they are terrible. I get about 30-50Mb/s rather than the advertised 1200Mb/s. I assume the signal has to go back to the distribution board to get between my adapters.

Take a look at orac decor. They have special baseboards for this

IMO a not-tiny house should have several sub-panels: Kitchen (enough to run the big stove + ~100A, but every outlet it's own 20A circuit), 'lab' / workshop (same idea), laundry room (lock-out breaker near / in room for connecting / disconnecting), etc.

A family member who's doing better in life than I am purchased a house built within the last 5 years, and it didn't even _have_ Ethernet run anywhere. WHY!?! I can't fathom how it's possible anyone would find that a negative.

> A family member who's doing better in life than I am purchased a house built within the last 5 years, and it didn't even _have_ Ethernet run anywhere. WHY!?! I can't fathom how it's possible anyone would find that a negative.

I'm less sure. Are there any 40 year old cables that you'd consider a positive in your house today? 20 year old cables? I wouldn't be too excited about having a house full of Coax, 10Base-T, and twisted pair.

I feel like the much better solution is to run conduit in which you can replace the cable rather than permanently putting in twice as much of whatever is in favor today. The conduit is always better, right?

Yeah, conduit makes sense. The future in homes - from this current vantage point - is likely all wireless via multiple APs around the house that have short range but very high bandwidth. Only us power users will have wired ethernet.

Wireless will _never_ have the bandwidth nor wall-penetration people want. If it has the latter all the neighboring houses / apartment units will bleed through, and if it doesn't even the limited frequency domain won't be enough unless you're in the same room as an AP that doesn't penetrate walls... at which point everyone will want drops running to the rooms they stay in anyway.

Yeah, conduits with pull strings and super-gentle curves (to safely satisfy minimum bends) are ideal... but at least run the lines there!

Re: Copper / Fiber. Fiber's a PITA for all users, even powerusers probably don't want it outside of the rack / enclosed areas. Delicate cables / connections.

5-6GHz is good at covering a house without too much leakage. And with 9 separate channels that each run at a gigabit or higher, you can have very good neighbor avoidance. That seems fine to me for covering both bandwidth and wall-penetration needs.

In a big apartment building you might have to use more narrow channels but I still think the overall picture for 5+6GHz is a good one. Especially because MIMO can squeeze a gigabit into 80MHz or theoretically even 40MHz.

The 6ghz spectrum finally looks useful compared to the DFS tainted (in practice, if you're anywhere near an airport or military air base NEVER use) 5ghz bands https://en.wikipedia.org/wiki/List_of_WLAN_channels#6_GHz_(8...

In 5ghz there was only a clean 80mhz + 160mhz block ; while the 6ghz full WiFi 6E / 7 devices can use has 3x full 320mhz (low power only) blocks. However as a negative, a _vanishingly_ small number of OpenWRT devices support either of those standards https://openwrt.org/toh/views/toh_available_16128_wlan6.0ghz and worse chances are really good someone's phone / laptop probably doesn't either.

99% of people don’t need the bandwidth. Of devices that can really benefit (mainly games consoles assuming your upstream connectivity is enough to cope), how many even have a copper port nowadays?

> 20 year old cables?

I just bought a new house and the previous owner had phone lines everywhere (bedrooms, kitchen, office, rec room - honestly kind of weird). The house was built in 2002 and I gather that it was common at that time to put in Cat5e and just not use all the pairs.

So all it took was a pleasant afternoon with the circuit toner, a new patch panel, and a few keystone jacks, and now the house is networked beautifully. Sure, they're not in exactly the places I'd have dropped 'em if I'd been starting from scratch, but it's hard to beat cables already in the wall for zero effort!

You are lucky they went with cat5. In my house there are phone lines everywhere, too, but real phone lines. Like in ISDN. That’s 8 strands plus some extra strands (the European ISDN bus needs 4 strands, so for a faked star topology you need 4 to every room and 4 back). It is neither shielded nor twisted of course. Just plain dumb copper wires. On some connections a stable 10 Mbit link is possible. From what I gather the cables were installed around 2000…

Five years ago I bought a house built in 2008, and every room has at least 2 ethernet (cat6 even, good for them), 2 coax, and 2 telephone. I guess this was some sort of transition period where people were starting to understand the benefit of running ethernet, but people hadn't cut their tethers from their landlines yet. The coax is a bit weird, though. Two ports per room? Seems like overkill, but I guess it's better to have and not need than the alternative...

I've still not found a use for the coax or telephone lines. I ditched cable TV 20 years ago, and haven't had a landline since 2004. The ethernet is super useful, though, for APs (paired with a PoE switch), and to to hard-wire the office, and stuff connected to the TV in the living room.

You can always use the coax for MoCa - but since you've already got ethernet, seems a bit pointless. I've got coax all over too, and, well, it's pretty much useless to me. But maybe someday I'll need a drop somewhere and be happy to have it.

Cat6 is twenty years old and still entirely usable today in a residential setting.

Even cat5 is not only usable but will probably give you 5gig or higher.

Dunno about m-gig, but I don’t like 10g copper as the power use is so extreme. I go for 1g copper or 10g fibre.

Yeah, I considered mentioning cat5 as well but figured cat6 was the better argument since people are still installing it today.

Have 20 years old cat5e. It actually gave me 10g

I was thankful I bought a house full of coax recently - now I have MoCA 2.5 adapters installed in the living room/kitchen/office/den/bedrooms all connecting to a central coax manifold which became my server room - this lets me hit ~2Gbps. Sure, cat6 would have been better, but that was 30 years from being invented when my house was built.

20 years is not that long ago.

30 years ago I was pulling cat5 into office building ceilings. Some of it is still there, still relevant. GigE runs just fine over it (most of it turned out to meet cat5e spec anyway), and very few applications on the desktop need more than gig.

Conduit is always better, but cat5 has proven to have a surprisingly long relevance.

> MO a not-tiny house should have several sub-panels: Kitchen (enough to run the big stove + ~100A, but every outlet it's own 20A circuit)

This seems like crazy overkill. Certainly there are kitchen appliances that devour power, but... what? Assuming an electric stove/oven, you really just need one dedicated circuit for that, and then everything else can run off another circuit. If you expect to use a microwave often, you might want it on its own circuit, but everything else you might have (blender, mixer, toaster, coffee/espresso machine, electric kettle, etc.) will run just fine on a single circuit.

And regardless, you don't need a sub-panel for this. All of these can run just fine as separate circuits off your main panel.

In usa at least in area where i live code requires dedicated circuits for most of kitchen appliances

If you have any way to pull cables between floors, try pulling DAC cables. They give you 10+G, require no splicing (hardware can be found cheap now), and generally just work as you'd expect.

With a 10G backbone, you can get away with basement->main and main->upstairs in a three-floor house, with a small bridge on your main floor.

> If you have any way to pull cables between floors, try pulling DAC cables.

Many of my devices take 10G ethernet, like MacMini and my MacBook Air docking station. Is there an easy way to convert from DAC to 10G Ethernet? The desktop has an SPF+ card so that will take DAC no problem.

Unfortunately there is no way to pull an any cables easily. But someone else suggested replaceable baseboards that hide cable runs, I should look into that.

I don’t know any media converter that would work for that but you can get a Thunderbolt SFP+ ethernet adapter. For example: https://www.sonnettech.com/product/solo10g-sfp-tb3/overview....

10G media converters are available for < $50 USD, but TBH I've never used one.

Why DAC cables over Cat6a or Cat7

They are the lowest power way to connect SFP. And they are cheap.

Though I'm not sure I would pull them as they tend to max out at around 6-7 meters. I use them to connect systems within the same rack/room.

Floor to floor in a house isn't too bad with 6-7m, but I suppose there's always the AOC cables that are 20m+.

10G Ethernet seems to be really hot, all my SPF+ connectors are hot, my TP-Link 10G switch is hot. It seems not be efficient anymore in that regards.

While my 10G fiber and 10G SPF+ DAC cables are perfectly cool.

I hope this is temporary but 10G Ethernet seems hot. I haven't seen anyone write on this topic yet but it is my findings.

It is well known that copper 10G runs hot. Use DAC or fiber if you want run cool

There is no reason to pull DAC when fiber has gotten cheap. Singlemode fiber is cheap enough now that isn't worth doing multimode fiber. You can get pre-terminated fiber that is easier to pulling whole DAC transceiver.

DAC is still useful for within rack.

True, though what I like about DAC is that there's far less risk of optical surface damage.

Sure there's ESD risk but I find that much easier to manage.

> running fibre in the house isn't practical yet

I think it is. A pair of switches and a single run isn’t that expensive and it’s truly beautiful.

But… I still wish I’d put in 2 cables and not one.

"(running fibre in the house isn't practical yet)"

What does that mean? I assume you are saying most builders don't offer fiber?

Running fiber in a house is entirely practical and easy to do. Cheap as well.

Might as well, but most end devices you’d want to connect would be copper. Most devices I physically connect are poe anyway. Sure my nas and desktop are fibre/10g, but my aps are all Poe copper, my laptop is copper unless I break out my dongle (although it’s rare ok not on wifi), my backup pihole is copper (I like having a second dns server in case the main vm on the nas fails)

For anyone else reading personally ceiling APs can't be overstated. My father-in-law is crazy about internet, but I've realized he was crazy like a fox when we did our to-the-studs renovation. He treated APs like a fire alarm or a CO monitor: one per room.

My brother-in-law thinks he's crazy, and obviously it only really works if it is practical to run the wire and you can afford the number, but for my relatively small house it is insane how little I think about my internet. It is one of those quality of life things I never would have realized with having someone come and do it for me.

I guess the only thing is some of the ceilings get a little crowded with lights + fire alarm + CO alarm + HVAC in/out + AP, so for small rooms it can be a challenge to figure out where to put things. But in that case I mean ... you don't need one per room. My father-in-law is crazy like a fox ... but he's also a little crazy like a crazy person.

These are great and common regrets in house wiring. I'm actually recording a video series on this exact topic right now. There are some things in electrical that are so easy and cheap to do but just get skipped over. Electricians are typically very very cheap (in a good for you the consumer way). However this tends to lead to under-building in a modern tech house.

Why 4 ethernet ports? You can just add switch later if that is the problem. I have this setup behind TV for Apple TV+TV

1. High-level, the post is all wrong. The point should be that you always need to make sure you can pull new cable. The poster illustrates this: single mode, multi mode, non-fiber, etc, etc. And if one "goes bad", you still can't run a new one, unless you have a pullstring.

2. The post cannot apply to fast/large networks - will be prohibitively expensive.

3. If running a few at home, I suggest to run MTP/MPO. It's basically a structured cable that can have around 12 fibers in them, plenty of future expansion.

Though I'll always run a large awg >>cat6 everywhere so it supports PoE++

what a bizarre reply.

the author is a sysadmin, who can definitely change what is plugged in to the switches their employer has chosen to buy, but doesn't have the power to make physical changes between multiple buildings, easily authorise opex spending on cable pullers or retrain as a cable puller.

please actually consider what you're replying to before pushing the REPLY button.

Pull strings don’t give you a way to pull new cables. They give you a backup way to pull new cables. Every cable in a conduit can be used to pull its own replacement or 2. But if anything goes wrong, you have a second chance before you have to go find the fishing rod.

> 2. The post cannot apply to fast/large networks - will be prohibitively expensive.

The expensive parts of fast/large networks are not the fiber strands.

* right of use / lease

* trenching and laying and covering

* amplifiers on long lines

* repair/maintenance

* endpoints

... which is why the first thing you upgrade are the endpoints, and the last you do is lay more fiber. Get the most you can afford (often physically) at the beginning.

> And if one "goes bad", you still can't run a new one, unless you have a pullstring.

The one that went bad is the pullstring.

> The one that went bad is the pullstring.

That works for short runs, but for long runs, the only way to pull the cable without breaking it is to use wire pulling lubricant.

And there's no guarantee that the cable you are pulling is undamaged. Rats don't improve pull strength, and even an electrically good cable can have the tensile strength cord severed.

If you are in that situation, pulling the old cable with dried lube on it may snap it in two, especially if it's pulling the new strand.

That's why leaving a pulling tape in each conduit is a good idea.

This is correct - I have pull strings in my conduits at my house for that reason. Pulling fiber down my 1000ft driveway would be difficult to do if the pull string was another fiber.

You can pull a new pull string in empty conduit pretty easily with a shop vac and a ziploc bag.

> 2. The post cannot apply to fast/large networks - will be prohibitively expensive.

Yes, it is obviously about small-scale sysadmin cabling. Telco networks have wildly different economics.

    > If running a few at home, I suggest to run ... that can have around 12 fibers in them, plenty of future expansion.

What kind of home network needs 12 fibers on a single run!?

Tautologically, a high-bandwidth one.

I ran fiber when I built my house a few years ago. I have MANY 12 fiber runs going to my IDF/MDFs, Server room, the gate, the dmarc, etc. I also ran SM fiber to every AP location. Out of 20 miles of pulls about 1 mile was fiber.

Why not copper with PoE to the APs? You still have to get power to AP somehow.

I have both. 2 Cat6As plus a SM fiber to each AP location. The APs need power over POE. I just added in my first AP that was 10g (Ubiquiti XGS), and that uses POE over 10g.

> 3. If running a few at home, I suggest to run MTP/MPO. It's basically a structured cable that can have around 12 fibers in them, plenty of future expansion.

For a homelab honestly just pull pre-terminated cable with LC connectors, a 20 or 25mm hollow pipe is enough. That way you don't need a splicer machine, experience on how to operate it, or measuring equipment.

If you have the choice, always run single-mode. It’s infinitely more scalable than multimode and the price is on-par. Around here, SMF is currently cheaper.

SMF optics are single digit dollars more expensive than MMF optics, if you use third-party modules. And you might as well use them because your SP does. In my experience they’re better built and more reliable (longer lasting and lower failure rate. n=1000s of modules) than first-party Cisco optics.

> In my experience they’re better built and more reliable (longer lasting and lower failure rate. n=1000s of modules) than first-party Cisco optics

And for the typical price of a Cisco optic you can buy two third-party optics. So now you have a spare. :)

From the last quote I got, I could buy almost 10 third-party Fs.com optics for the price of a Cisco OEM optic ($2,500 vs $250). And the FS.com seem to be manufactured more durably. And have a lower failure rate.

Personally I prefer Flexoptix for two reasons.

First, they are reprogrammable. They will send you a free programming box with your first order if you agree to a public testimonial on their website. Support are also very responsive if you need new codings they'll try to work with you to find something that works (they can push new codes to your coding box).

Second, more recently, they have the option to choose "not made in China". They sell both China and what they call "S2 Origin" which is any of Germany, Thailand, Taiwan, Vietnam and Malaysia.

Only downside is if you are buying for personal use you won't be able to buy from Flexoptics. They only sell to companies. (They do have a small number of resellers though, so you might have more luck with indirect sourcing for personal purchases ... you won't get a free programming box tough).

I'll check them out. That would be handy as I'm dealing with a bunch of vendors. Thanks for the head's up!

I mean, for a mere Benjamin, a disregarded LLC will solve that for you.

FS transceivers are also re-programmable: https://www.fs.com/products/96657.html

An addendum here that architectural firms and builders will still often spec out MMF for in-building fiber. You want to tell them "No, use SMF instead".

They will look at you funny and maybe even tell you "that's more expensive" because they haven't looked at price lists in over a decade. But they are wrong on all counts (unless you exist in a really whacky market). The only exception here is that some video, HVAC, etc systems require MMF to stay in warranty. In those cases, run both and use the MMF only for that junk.

No mention in the article of making sure to run conduit, so you can always add more cables easily.

This right here is important. We could be running fiber end to end in the near future. Ethernet would end up obsolete like the phone line.

How would running fiber make Ethernet obsolete? Physical medium and framing on the wire are fairly orthogonal.

People commonly use the word "Ethernet" to mean "Ethernet cable", as in a copper cable with an RJ45 on it, not just "Ethernet protocol", which is used over both copper and fiber alike.

It's interesting how long twisted-pair cabling has stuck around, to the point that it's associated as being the physical medium for Ethernet. Ethernet started with coaxial cables, but soon lost its association with those as twisted-pair was cheaper & easier to work with. And it's stayed that way for so long that a CAT-5 cable with RJ-45 (really RJ-38 but that's excessively pedantic) connectors is an "Ethernet cable", and a perfectly usable fiber optic cable is something else!

I've been in software for decades, and it was only right now that I realized that what I'm calling "ethernet cable" is maybe something else. Not that I understand the finer details of what you're saying.

You can find them all described here: https://en.wikipedia.org/wiki/Ethernet

Specifically these:

- https://en.wikipedia.org/wiki/10BASE5 (early 1980s)

- https://en.wikipedia.org/wiki/10BASE2 (mid 1980s, cheap installations like my high school in the early 1990s).

It is an Ethernet cable. It’s just not the only kind of Ethernet cable.

> RJ-45 (really RJ-38 but that's excessively pedantic) connectors

If we're being overly pedantic:

- the RJ numbers refer to not just the connector but the pinout and such

- the RJ (registered jack) numbers refer to just the "jack" side, not the "plug" side

So even though an "Ethernet" cable has plugs on it that fit into an RJ38X jack, its wrong as well to call it an RJ38 plug:

- different pinout

- the plug is the mechanically the same, but pedantically it isn't correct to call any plug a registered jack, because it's a plug not a jack

- the jack is mechanically different; an RJ38X jack shorts pin 1 to pin 4 and pin 5 to pin 8 when there's not a cable plugged in to it (this makes it a "series jack" instead of a regular "jack").

                        Name  | Mechanical jack                  | Mechanical plug
    47 CFR part 68 (registered jacks)
                        RJ45S | miniature 8-position keyed jack  | miniature 8-position keyed plug
                        RJ38X | miniature 8-position series jack | miniature 8-position plug
    ANSI/TIA-568 (Ethernet)
                        T568A | miniature 8-position jack        | miniature 8-position plug
                        T568B | miniature 8-position jack        | miniature 8-position plug

I don't believe any registered jack uses a miniature 8-position unkeyed jack (but scanning through 150 pages is a pain and for some reason Ctrl-F isn't working reliably in this PDF).

(ANSI/TIA/IEC calls them "modular" jacks/plugs, while the FCC calls them "miniature" jacks/plugs, but same thing. Well, same thing the way that CAT-5 and CAT-5e are the same; same, but with different specified tolerances. And just saying "Ethernet" or "568" isn't specific enough here; CAT-6 requires an IEC 60603-7-4 connector, while CAT-5e only requires an IEC 60603-7-2 connector; even though most people would call those the same, they have different tolerances. I don't care to compare the FCC's 47 CFR part 68 requirements with the IEC 60603-7-{2,4} requirements.)

> (really RJ-38 but that's excessively pedantic)

RJ-38 is a 8P4C (eight pin, four conductor) modular connector with shorting bars used to allow an alarm system or similar to seize a phone line when plugged in but for the line to still operate normally with the plug removed.

You might be thinking of RJ-48 which is used in T1 service and thus often has "RJ45" cables plugged in to it, but technically that's 8P4C as well, also with (differently configured) shorting bars that will physically loop the T1 if disconnected.

Ethernet uses an 8P8C modular connector but it's not any of the Registered Jack standards

It should mention conduit for sure, but sometimes conduit is hard if you want to stick to the rule of no more than 180 degrees of bend in the line (and no more than 90 at one bend). Long conduit runs need junction boxes at some of of the bends, and it's hard to always make those accessible.

I thought you were allowed 4 90 degree bends, or 360 degrees between boxes.

Code allows up to 360 degrees, but it gets really hard to pull after 180.

Well, conduit plus multi-cell innerduct to make it easy.

First I've heard of this! What's the advantage, in your experience?

It’s easier to pull new cables in a conduit through an empty cell of innerduct than it is to pull a new cable through a conduit with existing cables already in it. It tends to be used for fiber conduits.

Here’s a product I’ve purchased for my electricians to install: https://www.maxcell.us/shared-downloads/cp145_MaxCell_Edge_D...

Fascinating! Thanks for the explanation and the example.

I recall working in a building that:

1. Was the former HQ for a local telecom company 2. Was now an office building with a couple floors of data.

Ages ago, one of the DC providers in the building had run 36 cores to the roof to service a telco.

It was a massive undertaking, the building is riddled with asbestos and old plant. Had multiple renovations etc.

Anyway, a few years later, every telco in the state wanted to be on that rooftop. And they had the only reliable means to service it with fibre. That 36 core became a massive profit generating asset, one that they could have monetised even further if they had have run 100 core instead. But it was never feasible to drill holes for a new duct (Asbestos regulations getting tighter), and it was never feasible to shut down 12 telcos for a day to use the existing fibre as a draw cable.

Getting a single core rented here, and throwing on Bidi's, was like mana from the gods at the time.

I just had this experience. I had a 200ft run between my house and barn. The original builder put a direct bury ethernet between the two and it failed. I dug a trench, put in a conduit, pulled 2 fibre lines and left a pull string in.

I recently had the primary fibre fail and am now on the backup. If I need to pull new ones in the future I can do that pretty easily through the conduit.

Yep. Direct-burial ethernet is surprisingly vulnerable to nearby lightning strikes. It's not a matter of IF the cable or devices get damaged, it's a matter of when. Nearby (not even direct) lightning induces ground voltage potentials between buildings to the tune of hundreds of volts or more.

Do you have experience or information on direct burial ethernet for something like a POE camera? I'd like to put one on the back fence to watch the back of the house and yard. Direct burial in the back yard would be a plenty easy thing to do, but the cable is pricey enough that I've held off for now.

Whatever you do make sure you put a fiber media converter between the “ethernet” cable and your network to prevent your entire network getting fried.

I used to consult for an ISP that put direct bury in a region where it snows 1/3rd of the year.

It was profit generating. They would offer to put conduit in for an extra fee, the customers always said no, then they would be back to install conduit and cable in the spring after the ice had killed the cable.

    > I recently had the primary fibre fail

Woah, this is surprising. Do you know the root cause? I could imagine copper cabling is much more sensitive to the outdoors, that is why I am so surprised about your fibre failure.

My business is fiber broadband, and we are laying as much as we can! Fortunately the legislative environment in the UK is conducive to this goal. For now we provide XGS PON (10Gbs minus overhead), with 50G PON becoming available in some locations later this year. Unlike regular point-to-point, it is point-to-multipoint through the use of optical splitters with one head-end port capable of servicing up to 256 clients. With the UK being so far behind most other developed countries, it means it can take advances in all of the R&D from the last twenty years. It has also led to the strange situation where there are some properties that have a choice between 4 or even more fibre network providers. Some fiber networks are tied to a single ISP whereas others are open to wholesale. I am not sure if there are any ISP's that support multiple connections to the same property via different fiber networks - but I think you would generally want to use a different ISP for each network anyway. IPv6 is supposed to make that easier, though in practice there are still no protocols (a la MEF 17 Service OAM) for the ONT to signal to the router that there is a fault condition. This is one circumstance where PPPoE is useful. As the UK has been slow to deploy fiber there are many ISPs still using it. I found this out the hard way when I made the switch. My trusty old Ubiquiti Edgerouter Lite (Cavium silicon) is not capable of hardware offload for ipv6+pppoe+vlan at the same time, so anything particularly demanding on IPv6 (e.g. Steam) essentially locks up the router. I will be trying out an Alta Labs Route10 this week to see if it provides any improvement.

Not as a serious other-side but I knew an engineer associated with ARUP and during the time of Canary Wharf being in design, the plenum spaces were re-sized twice as a function of emerging pre-fibre burdens.

I believe now, there is probably the equivalent of 3 or more floors of unused space, consumed by oversize plenum in the light of emerging fibre.

I do believe fibre probably is a sensible end road for "small enough" and so pulling more of it, for the same radial bend and cable diameter probably makes sense.

Also bear in mind "re clocking" by tuning the lambdas and deploying new lasers is a thing. That 100 pair fibre can be re-clocked across it's lifetime and be the equivalent of 1000 pairs in equivalent bandwidth, by the time you're done.

“We’re only going to do this once and there’s always the unforeseen.”

— Joseph Bazalgette, on doubling the pipe diameter when building the London sewer system

(though I don't have an original source for that quote)

I'm surprised Chris left out a big reason: cable is cheap, and labor is expensive. And there are a lot of fixed costs that don't depend on the number of cables or their thickness.

You might save a little bit going with 4 pairs instead of 24. But that goes out the window as soon as you need to run a single new cable. If you want to be stingy, pull the cables but leave them unterminated.

We've almost come full circle with the idea of "run more cabling" already in some ways.

For example, a building built in 1960 had analog phone lines. Then in the 80s network lines were added. Then in the 90s - 00s more and more and more network were added.

Then in the 2010-2020's we're starting to wind back down. Removing switches and racks that used to be fully populated with CAT5 which are now mostly empty. The end devices that needed these runs are now running on WiFi.

To elaborate a little bit;

I've been in many, many buildings for cabling issues. While I generally agree with your assessment of "run more everything" I think this could be taken out of hand for a lot of smaller IT departments who aren't capable of using more advanced methods of network building. Like subnetting, VLAN trunking, or redundant links.

Why would you need to learn about these things if you've got 3 or 4 more pairs of fiber hanging down? Then you have a situation that I've seen before, where you have 2 fiber runs carrying flat networks right next to each other with just a couple of IP phones on each one.

> Some of the time this fiber failure is (probably) because a raccoon got into your machine room

Or into your handhole/vault, along with her babies:

https://old.reddit.com/r/FiberOptics/comments/1ji3rrt/its_no...

Regarding single mode vs multi-mode ... other than cost, is there any reason at all to run multi-mode fiber in 2025? I was under the impression that single mode was better in basically every way.

Multi-mode transceivers used to be a lot cheaper than single mode, but now they're around the same price for short distances. From what I can tell, most datacenters appear to be making the switch to single mode even for in-room networking.

Multimode transceivers are still a slight bit cheaper ($30 for 10GBASE-LR vs $22 for 10GBASE-SR on fs.com) but that's more than offset by the single mode fiber being slightly cheaper, even for short distances.

Apples to oranges. SR is for short runs inside a building. LR is for sending signals across a city.

For short runs inside a building, the transceiver prices are much more comparable. Compare https://www.fs.com/products/251819.html?now_cid=4089 to https://www.fs.com/products/282973.html?now_cid=4089

SR is for when you accidentally installed MMF really.

Except, no, you can run (single lambda, not muxed) LR transceivers on MMF, it's out if spec but you actually get around +20% range before hitting dispersion limits.

I'll also note SR ranges are not in fact long enough for larger buildings. And this gets worse with higher speeds, since modal dispersion is a bandwidth problem, not attenuation.

Also, using 800G for comparison is a bit ludicrous ;). 10G is bread&butter, 25G/100G is mainline. (40G is the shunned weirdo uncle.)

Historically, the price difference between MMF and SMF transceivers was greatest at the highest speeds. That doesn't seem to be the case anymore for short-range transceivers.

Why does multi mode sound like it should carry more bandwidth but from context seems to be the less interesting option?

Mode is simply the wavepattern of the light traveling through the optical fiber. Multimode means there are a few different wavepatterns traveling down the fiber, they tend to potentially interfere and its not as efficient of a use of optical power. In singlemode, this is a lot harder to do and requires better optics, the light is in one pattern, power is efficiently allocated to that pattern, and thus the light can be sent much further. Its been many years since college so hopefully that explanation suffices.

Now lets say you want to send multiple signals to get more bandwidth on one fiber, you just need to move the frequency/wavelength of light so different signals have different wavelengths and can be discerned at the receiving end. That's gonna take even more optics and filters on both ends, but works quite well to add more bandwidth to existing fiber installations where running more fiber would cost a lot more than installing new equipment on the ends.

Rather than interfering with each other multimode suffers mainly from modal dispersion as the modes do not travel the same path and some take a shorter or longer one, so the signal is spread out.

It's DAC in rack and then fiber or copper between spine and leaf (depends on distance, etc).

Cost, and marginally, heat of the transceivers.

    > heat of the transceivers

I never heard this before, but I have no reason to doubt you. Are multi-mode or single-mode transceivers hotter, and why?

Single mode uses more power because the light emitter is more powerful, usually

No. Just run single mode.

Originally thought this was excessive, but at least 4 per outlet/room. Saves buying a bunch of smaller switches.

1 for a switch (Although VLans can help) 1 for telephony if required 1 for media devices / peripherals 1 extra

Optional fun use of HDMI over ethernet for more runs.

Behind TVs, you can break it down, or just run a bunch of runs.

I routinely wish someone had designed a power outlet shaped switch that runs on POE.

But some brands would run too hot for an enclosed space and a Netgear one would burn your fucking house down.

I've seen some 'in-wall' access points that might be kind of what you're looking for. Designed for hotels, etc, they have an ethernet port on the back with poe-in, and usually two ethernet ports on the front, and also do wifi.

I have no experience with them, though.

There are modular jacks that can expose 6 or 8 different connectors on one outlet. I just figured the hotel ones were a slightly more sophisticated version of that.

I've been pretty happy with Unifi's "In-Wall" AP (e.g. https://store.ui.com/us/en/category/wifi-wall/products/u6-iw): PoE powered, has four downstream RJ45 ports (one with PoE itself), and is a side-firing WiFi AP as well. Like the sibling comment said, they're made for hotel rooms, but I've never been in a hotel fancy enough to use them (possibly because the RJ45 sockets aren't particularly discoverable, being on the bottom). It doesn't live in a power outlet, but it's meant to be mounted at the same height. Worked great for my home office. (I'm not associated with Unifi, just happy with their gear)

Yeah surface mounts aren't bad. Combined switch and Wifi is good.

Seems like for discoverability you'd have to mount it backward or upside down somewhere so that people can see the ports, and then what about the wifi signal? I see what you mean.

There are some depending on the number of ports you're after. I'm not sure if it would run directly on Poe, but maybe PoE could be converted to USB power to run something like a GL-iNET device or something that sized.

https://www.gl-inet.com/products/gl-b3000/

A bigger issue with ethernet presently and for the last while is unreasonable prices on 10GbE both adapters, switches, etc.

Sticking with other mediums for now for 10GbE

In a home buildout, running fiber is surprisingly inexpensive and easy to do. There is one area to watch out for - In most US homes drywall is installed with screws that go into studs so if you have large bundles of wire or fiber be careful how the sit relative to the surface of the studs. Often the drywall screws will be off by a bit an miss the stud. If you wire bundle is large enough and it is a 2x4 wall vs a 2x6 wall you can lose a number of wires with a puncture.

A second common problem is the wire installer pulling the fiber around sharp corners, sometimes with staples and zip ties. Fiber is surprisingly resilient, but very sharp bends will cause significant problems. Using armored fiber can help this a bit.

> Using armored fiber can help this a bit

You can also get bend-resilient unarmored fibre. Its what the incumbent telcos use when they send out their monkeys (sorry "engineers") who've been on a 4-hour afternoon course on how to install fibre before being let loose. Old copper "that'll do" install habits die hard in the telco world.

Nail plates.

Screws should _never_ be able to hit the wiring. If that's possible, someone failed to install nail plates or Smarshield sleeves.

As for the sharp corners, IMHO you should always install a raceway of some sort (everyone just uses smurf tube), then pull the fiber in afterward. It takes real effort to fuck up the bend radius of conduit.

I was not referring to wiring going thru studs, but along studs. Nail plates are very effective for going thru studs, but often wire is ran along the side of a stud. If it is a large bundle and you don't keep it centered on the side of the stud it will be close to the surface right where a screw could miss. Of course putting some kind of shielding or sleeve could work, but that is very uncommon in house building.

Smurf tubes work great, but are not practical in some high density environments, especially a house with steel structure and I-beam joists. I used these for some locations in my house, but not all.

I left one small room in my new home without network cabling, finished just before covid. Guess where my home office ended up being located...

https://michael.stapelberg.ch/posts/2020-08-09-fiber-link-ho...

King County (Seattle) did this when wiring up the municipal network. Networking World ran at least two cover stories on this in the mid-1990s. They ran 200 ... bundles? Strands? Everywhere. That fiber is still there, under sidewalks, inside bridge structures, everywhere. Whoever bought the rights to that excess fiber from the city is probably subletting it to various FAANG and datacenters making a small fortune. That fiber runs through most of downtown seattle.

> Fiber comes in two varieties, single mode and multi-mode. I don't know enough to know if you should make a point of running both[…]

Let me fill this in: in 2025 you (and everybody else) should be running SMF. If you need to directly connect to existing MMF you should run that additionally. But do not build MMF-only in 2025. It's akin to installing Cat5 (non-E) or even Cat3. And you pay mostly for the work, not the cable. Do yourself a favour and put SMF in.

Agreed. SMF can handle nearly anything. A single strand can handle 100Gb 10km for OS1 or 100km for OS2 per frequency and you can do multiple frequencies at the same time or go bidirectional, too.

I subscribe to this philosophy.

I recently had my house remodeled. A lot of the walls were opened up and I asked the contractor to run fiber and cat 6a everywhere. Now every room has multiple drops of each. Even the outdoor and garage have several drops. Bathrooms have ethernet. Attic, basement and storage room has multiple drops. I may have overdone it. But, I never wanted to every worry about this ever again.

I thought about multiple drops of cat6 per room in my house, but they charged me per drop. So I just put one drop in each room and added a switch in a room that needs multiple connections. It's usually for Wifi WAP, smart TV, and the Tivo box.

Nice story. Why did you choose to drop both fiber and copper? My assumption: You can run 1/2.5G very cheap these days over copper. When 10G/25G gets much cheaper later, you can switch to fiber.

There is no power over fiber.

I was under the impression that once laid, fiber is the most stable and future proof between all mediums (barring physical damage). After all it won't rot, stain or degrade like copper. It's plastic, mostly.

It is, it's just not immune to physical damage. Rodents and Jackhammers…

Backhoes can find fiber from miles away... they hunger for it

Depends on timescale.

Like if you are in an area that has frequent natural disasters or just heaps of construction work, you will find that RF is more stable and future proof.

Also, you can get issues with cable bends and terminations. Which are simple to fix, but often go undetected by cheap fiber carriers.

I manage a couple pair of dark fibre for a customer, and the fibre runs through a couple of suburbs that are under constant development. And they have near monthly half day outages, and quarterly 1 - 2 day fibre outages.

Like others have said, backhoes hunger for the glass.

And if you can't, always pull a string through the conduit .....

Just always pull a string through the conduit, full stop.

As the article says, this is true for just about any cabling, copper, power, etc.

I’ve done cabling in two houses and I’ve never had too much. I’m always finding reasons to run more to new places.

Had a squirrel chew through the fiber internet cable going to our basement. Fortunately there was enough slack that the unchewed cable could still be pulled and reach the telephone pole and reconnect without having to redig the entire path, which would have cost much more.

At home I have a pair of single mode fiber running in each room, it is future proof enough for me.

ATM running at 1Gbps, but I will soon upgrade the router, switches, SFP+ modules to run at 10Gbps since my internet provider will upgrade the offering from 2.3 Gbps to 10 Gbps this year.

YAGNI: You Are Gonna Need It.

Reply on news.ycombinator.com