GFiber Labs announces first project: 20 Gig with Wi-Fi 7 | Hacker News Reader

GFiber Labs announces first project: 20 Gig with Wi-Fi 7

fiber.googleblog.com

50 pointsxnx68 comments

I'm still salty about Google Fiber. So much potential and Google just... gave up on it? Google giving up on things is basically a meme now, but back then I couldn't really understand why. Customers raved about it, everyone loved it. You could see the entrenched ISPs in an area magically drop their prices whenever Google Fiber came into town. Then they started facing regulatory hurdles thrown by the incumbents and it felt (or I felt) at that point that they just didn't want the fight. I couldn't understand, this absolute behemoth of a company, just laying down and giving up.

It seemed like they'd given up in 2016 with the announcement to stop expansion. But it seems like they've got the ball rolling again, probably at a slower pace. They announced last year that they're expanding into 22 new areas.

Recently here in Charlotte NC they have rolled out access to 5 and 8 Gbps connections. More than I have need for (I don't want to upgrade to > 1 Gbps networking haha), but great to see the investment.

Is that symmetric?

Most of their plans (1, 5, 8) have symmetrical speeds. The "2 Gig" plan is asymmetrical with 2 Gbps down and 1 Gbps up.

I’m kind of in the same boat. I average around 1.2Gbps and I’m not sure why I would upgrade to anything faster.

I always thought of things like Fiber, Fi, Auto, etc as more "pressure" projects. AT&T is the only service provider in the area and charging $50/mo for up to 20mbps download? Bring in Fiber, and within a few weeks of announcing AT&T introduces a 1Gig plan for $60/mo. I watched this happen in the wild. But the thing here is that anyone being on the internet benefits Google because how much control they have over it. Same goes for phones/android. Auto increases maps (even 3rd parties buying data from Google maps and using it in their own map system).

But it is bittersweet. Because it only works if people buy into the google product but then google often does a rugpull. I think most people are probably better off but there are definitely casualties.

If you think about why Google was doing Google Fiber, it was to increase users' bandwidth, not really to take over home internet service. Their announcement of city selection resulted in incumbents in those selected areas doing rollouts even before Google could get site selection done. At that point, the mission was accomplished, and Google didn't need to do anything.

But, they can't really do that trick a second time, so everybody else is slowly waiting for the existing lines in the ground / on the poles to deteriorate to the point where incumbents have to rollout new stuff.

But at the same time, they always have struggled in finding alternative verticals to make money from, outside of search and ads. That’s why I always thought their end goal with Fiber was to own the pipe and whatever else was at the other end. They could have become a major ISP in the USA.

But broadband can't generate Googly margins and they failed to find breakthroughs to reduce fiber costs.

Cloud also can't generate Googly margins but they're willing to invest in that.

Google already works at a cloud level. I wouldn't be surprised if Google cloud isn't part R&D for internal systems.

Cloud margins are pretty big. Or they should be given what they charge.

I think the costs were just too much to expand as rapidly as they wanted. They experimented with "nanotrenching" in Louisville, burying cable directly in 2-inch-deep channels directly in the road surface, which ended up in disaster [1]; shortly after, they "paused" expansion, fired their CEO, and went through layoffs.

1: https://gizmodo.com/when-google-fiber-abandons-your-city-as-...

I lived in Louisville when this happened, and met the guy who came in second place for Louisville portion of Google Fiber.

1. Yes, the microtrench became a disaster, because the material they used to cover the drench stuck to people's tires when it was hot outside and it pulled up. Paving over it would have solved the problem

2. Google chose to hire a lawyer to run the project. The second place guy had installed fiber his entire life, had personal relationships with the people that owned the right of ways to connection poles (ISP). The ISP folks did not like the lawyer and did not work with him and that's why it really died. (according to mr. second place)

> Yes, the microtrench became a disaster, because the material they used to cover the drench stuck to people's tires when it was hot outside and it pulled up. Paving over it would have solved the problem

Paving it over would mean digging the trench deeper, meaning it's no longer a "nanotrench" (Google's experiment) but a "microtrench" (a very popular method of fiber deployment). These are trenches that are cut down below the road surface (12-24 inches), filled in, and paved/sealed on top to match the road surface. For this you need conduit, filler material, asphalt sealant, and a bit more labor.

That was the point they decided to give up on wires and invested in SpaceX with the intent of blasting the internet into space. Like.. that'd be faster than dealing with the government at every single pole within disputed territory like they had been. Getting the incumbent to come out at a pre-agreed time to be there and check the site before they worked or assist, for example. So they took a moonshot instead.

Seems like its starting to work?

I have had Google Fiber in Austin for years and it has been horrible.

They farm so much out to third party contractors who have no idea what they are doing.

I've had so much downtime (over a month total in the last 2 years) and support just sends the same couple of guys out every time. The first time he showed up, he said that the LED on the front of the router was dim (which it wasn't, exact same from day 1) which indicates it is faulty. Not only does this make no sense, but they have replaced it 4 times with no luck.

They downgraded my plan from 2gb to 1gb to "fix" the problem which stopped the constant downtime, but significantly downgraded the quality of the connection and massively increased my latency. I've tried to get them to do anything beyond just replacing my router over and over again, but they just keep sending me to another automated system which sends out the same guy who thinks the LED being dim is the problem.

I had ATT Fiber for years before this without a single issue. And because Google Fiber basically bought out the neighborhood I have only one provider option now.

Just like everything else Google does, there is horrible support, it is as automated as possible, and if you have any kind of problem you're basically out of luck. When it works it works, but anyone who has ever had trouble with a Google product knows exactly what I'm talking about.

When they announce stuff like 20gb connections I can't help but laugh at how insane it is considering they can't get a reliable 2gb connection to one of their pilot cities they've invested billions into.

It is kind of insane that I had to create a new google account just to change addresses... to another address that had actually had google fiber two years PRIOR to my existing apt...

It's insane to me that a company like Google couldn't figure this out. Also, when I created the new account I was locked out of google payments for 2 days because they thought someone was pretending to be me (after I made the new account). Huge nightmare but at least my 2G fiber works.

I have gfiber in Austin and it's fantastic, and if I really wanted to I could still get ATT fiber or spectrum cable.

How did gifber "buy out the neighborhood"?

I've been using it for years and pretty happy, but: have you tried just replacing the modem and router with your own kit? Last I remember, IPv6 is slightly fiddly, and you need some VLAN tagging set up, but you can just use any DOCSIS modem and a bog-standard x86 box to do routing. I ran pfsense for several years that way and it was great until the Atom chipset died (infamous C2000 series bug.)

Even then I totally sympathize that the standard equipment should Just Work and I understand that babysitting networking gear isn't fun. But if you actually can't get anyone else, it might be worth a shot.

DOCSIS is for cable, not fiber. Fiber equipment is usually ISP-specific - I'm pretty sure I can't replace my AT&T ONT (even if I can replace the gateway in a hacky way).

I thought DOCSIS was cable-only. Most fiber ISPs (PON) require an ONT, and they'll give you an ONT + Gateway or a Gateway with an integrated ONT (the former lets you use your own equipment by plugging directly into the ONT).

ISPs that offer dedicated fiber lines let you use an SFP with your own switch/gateway directly, if I'm not mistaken.

>ISPs that offer dedicated fiber lines let you use an SFP with your own switch/gateway directly, if I'm not mistaken.

At least with Comcast and AT&T metro ethernet, they provide the modem/gateway - you cannot terminate in your own equipment. You get to use your own router, of course, but they will never let you directly touch the fiber. Don't like it? They're the only metro-E providers in town (or rather, a specific area of Nashville, TN I got to see quotes from), so if you need a dedicated line and don't like their rules, you can kick rocks unless you've got money to burn. Comcast charges $12k to install, but that's still pennies compared to paying a tier 1 that doesn't already have a presence in the area to lay a line to your door.

You don't even need to do that anymore. You can just enable BYOD in your settings and plug in your router. Just be sure to get one that can actually handle the throughput of the service, most of the really cheap ones can't.

In my experience an FCC complaint will get that sorted really quickly.

https://consumercomplaints.fcc.gov/hc/en-us/articles/1150022...

Do you have a combined fiber modem or modem->router. ISP buy shitty hardware don't always retire it when it becomes problematic. There is basically no good reason to rent ISP hardware.

My first connection was a 28.8kbps (advertised) modem, so I am very appreciative of my current connection speed, but I'm still very jealous of those with access to Google Fiber and a 240x faster connection than I have now.

It's not all sunshine and rainbows though. I've been lucky to have Google Fiber for a number of years and I still get buffering issues with YouTube and YouTube TV.

I love my google fiber. AT&T forced me to use a massive, shitty all-in-one modem router wifi access point when my fiber was through them. You could put it into passthrough mode, but if it reset for any reason you’d lose your config.

Google just gives you a small wall component and an IP address. Whether you use their router is your business. It’s fantastic.

This is how all reputable ISPs operate. My Sonic service just terminates at the wall. They will provide equipment but you can equally well bring your own.

We provide equipment because 99% of our customers don't want to deal with it. They just want wifi and a decent connection that isn't DSL speeds. I will do a passthrough port if someone really wants it but you are getting an ONT regardless so I can signal test and get outage notifications.

I didn't mean to suggest that customers would bring their own ONT. I've never heard of people doing that.

We don't always use an ONT, sometimes we use a enterprise switch if it is an odd deployment, but it lives on our side of the network and the customer gets a port to use.

When I had ATT I told them my gateway was broken and to send a tech out. When they showed up I asked for a dedicated ONT setup, which he was happy to do and asked for a tour of my homelab.

Maybe I was lucky and they don't carry that equipment around anymore.

For anyone unfamiliar, PON means that several users will be physically spliced onto one shared fibre. In most of the deployments I'm aware of, it's 32 users per fibre. It also requires specific hardware, not standard off-the-shelf SFP modules like normal FTTH deployments.

I suggest "20 Gig" with PON should come with a massive asterisk; when you read "PON", you should think "Comcast" because you're about as likely to get the advertised speed from PON as you are from Comcast.

> It also requires specific hardware, not standard off-the-shelf SFP modules like normal FTTH deployments.

What is a "normal FTTH" deployment? What percentage of residential deployments are PON versus non-PON? Because I was under the impression that PON was the (vast) majority of residential fibre deployments, in houses and apartments/condos.

And by 'standard OTS SFTP' do you mean Ethernet? Because the IEEE has their own PON standards, but also once you get outside the data centre, office building/tower, or campus, there are all sorts of 'different' technologies you tend to have to deal with WANs depending on distance.

We recently switched ISPs at work, and our 10Gb link was a BiDi single-mode fibre and we were provided with a media convertor from which we connected our router-firewall. Depending on distance desired you may (>80km) have to start getting into coherent equipment or optics (CWDM, -ZR).

>It also requires specific hardware, not standard off-the-shelf SFP modules like normal FTTH deployments.

Sorry this sentence doesn't make any sense unless it is US specific. There no such thing as "standard off-the-shelf SFP modules like normal FTTH deployments". And normal FTTH deployment around the is PON. Or specifically GPON.

Every single residential internet connection is sharing bandwidth one way or another. You are not owning a piece of the internet backbone.

I don't know of many places splicing in splitters. You can do it in a distributed split but distributed splits are a nightmare to maintain. We home run all endpoints to splitter cabinets to a patch panel. Then we patch the input fibers into splitters and then splitter output to house. This is all in a cabinet that a tech can walk up to and work on without tools.

> I don't know of many places splicing in splitters. You can do it in a distributed split but distributed splits are a nightmare to maintain.

Could this be dependent on geography?

Perhaps distributed (cascaded?) splits may be useful in less urban / more rural deployments where stringing a large number of strands over very long distances can be more pricey? Perhaps better to have x strands strung and patch in at any poles that are near residences?

* https://www.commscope.com/globalassets/digizuite/915282-taps...

* https://www.commscope.com/globalassets/digizuite/3101-optica...

* https://community.fs.com/article/centralized-splitting-vs-di...

We are pretty darn rural, not quite kansas rural. I guess the way we look at it is that the cable doesnt really cost anything in the grand scheme of things. Getting it built and maintenance really dwarfs the cable itself over time. Plus it allows you to do piecemeal upgrades to XGSPON instead of having to do an entire distributed split all at the same time. We can feed a splitter cabinet an XGS optic and patch each individual customer over as they upgrade. It just provides some better options at the cost of wasted fibers.

You're always sharing bandwidth. It's not like you have committed capacity end-to-end to every site on the Internet.

As much as I hate Comcast, I never had a problem getting the advertised speeds from them.

I'm on AT&T PON and I get higher than I pay for (I pay for 1Gbps and regularly get 1.2Gbps+ through a 5GBase-T link). It helps that I live in a small town and my neighbors aren't very noisy (in a networking sense)

What companies do "normal" FTTH deployments? My only other experience is with Quantum (née CenturyLink, née Quest), who definitely use GPON.

I know of a few coops that did active ethernet for future proofing after they got badly bitten by DSL. I doubt they even got that close to paying for the DSL rollout before they ended up ripping it all out for fiber.

Ugh, I see PON, my eyes roll. I know it's all the rage for fiber providers since it's cheap, but it's a cop-out and a cheap-out. All new fiber deploys should have home runs. Fiber is so incrediblyt cheap it's stupid not to do so. Otherwise you'll just be stuck with the idiocy of having to manage the worst of both cable in sharing bandwidth in some situations and WDM in having to manage x8 or x16 or even x32 splits and associated power budgets. It's just a lose-lose, even if it's 25G!!!111.

Build it right or don't do it at all.

This is a hilariously bad take.

PON works well for residential and SMB last-mile connectivity. Especially compared to the shitshow that is coaxial troubleshooting.

>This is a hilariously bad take.

I am starting to think if I should completely give up Hardware topics on HN. 4G/5G ( 3GPP Rel ), WiFi ( 802.11 ), USB, Thunderbolt, CPU, GPU, FPGA, DSP, Foundry, PON, DOCSIS, DRAM / NAND, Display......

It is getting ridiculous the amount of noise Vs signal in these topic. The only good thing left is at least we still have enough people coming out to correct all the Anti-GPON misinformation.

> Build it right or don't do it at all.

Or build what you can afford/finance now, and expand later.

There's no technical reason why active fibre can't first be deployed to the node or curb or top of the street first, with PON going down the rest of block/street to begin with. This allows for must of the gains over co-ax and especially DSL.

If Internet use cases ever reach the point where an entire street or floor of an apartment/condo is contending for the full bandwidth of a 25Gb PON fibre run, they can then perhaps split from (say) 1:32 to 1:16, 1:8, etc until each customer gets an individual run.

There's only a finite amount of capital, and spending for individual runs for early deployments may delay other people.

IMHO it will be a number of years before people can regularly saturate >1Gbps, and much longer before 10Gb use-cases start appearing.

As it stands HS-PON (ITU G.9804) can do 50/50, 50/25, or 50/12.5:

* https://en.wikipedia.org/wiki/Higher_Speed_PON

50Gb split by 32, is 1.5625 per customer before contention starts. Are there particular use-cases where a sustained 1.5Gbps (up and/or down) is not sufficient in 2023? Of course it may be possible to burst higher if the rest of the fibre isn't busy. I currently have 1Gb, and software updates certainly come down faster than my previous 25Mb VDSL, and 4K YT videos don't buffer as much. Though I'm not sure if things would be materially different than if I had (say) 'only' 500Mbps.

I am not sure the people bashing GPON know how a typical network is constructed. If GPON doesnt work out, we dont ahve to rip it all out and do it again. There are home runs back to all of the splitter cabinets. If we need to switch to home run back to the CO, we could do it fairly easily. We just run larger cables to the cabinets and patch directly across instead of patching input to splitter to house.

I understand how they are constructed, I've supported them in the past. What I quarrel with is the splitter aspect and the additional equipment that inevitably appears between the subscriber and the final termination point--be that a splitter, be that yet another patch that could fail inside another cabinet along the line in the middle of nowhere.

If you're going to make p2p home runs, just do it. Support the equipment in the first place at the CO to make that possible, rather than installing splitters then ripping out those splitters, patching things back up. Remove the intermediate cabinets. At that point you may as well just provide raw ethernet. What was the point of the PON nonsense then?

PON is a stop-gap technology that's outlived its usefulness, if it ever was there before. It's a crutch and should be retired.

PON is what's allowed fiber deployments out for consumer affordable prices and continued scalability of endpoint bandwidth, so "stop-gap" seems a little bit of an extreme position. There are a lot of environments where just the space and power requirements of a 10G port for each customer make it a non-starter, let alone the cost.

> PON is a stop-gap technology that's outlived its usefulness, if it ever was there before. It's a crutch and should be retired.

Considering I just moved from 25Mb VDSL to 1Gb GPON in the last few months I find it very useful: certainly my software updates come down much more quickly (though my SSH CLI sessions for work as a sysadmin aren't really that different).

I'm not sure what I'd find more useful about a direct connection over a shared-medium fibre.

If GPON or XGS ends up being a mistake, it can be fixed pretty easily as long as they didn't do distributed splits. As it stands currently, GPON itself is very much overkill in our deployment.

I wonder what the CPE looks like. It must be an integrated ONT+WiFi right? Getting 20gbps from an ONT into a router would be a challenge for UTP.

Can't wait to not get it in Nashville, except for a couple of buildings.

Can't wait for them to bring this to Austin!