Hacker News Reader

> Living in a home with Cat5 throughout so the best I can do is to route 1Gbit.

Having your uplink faster than all your individual ports in a network helps prevent any one port from being able to saturate your network.

I work at an ISP that within the last year started selling 10Gbit symmetric, and we all sorta know there's not a huge use for it - yet. It's not really an issue for use to justify though, we don't charge any extra for it, it's all $40/mo and if you're in a new area we're building you get 10Gbit instead of the old 1Gbit at that price.

Slightly off topic, but have you heard of Utopia? They’re a municipality in Utah. They run the fiber for a flat fee to customers and let any ISPs sell on it. Currently ~10-15 ISPs. Your guys pricing would absolutely stomp these ISPs. Have y’all considered jumping on these kinds of opportunities? No need to run cable and can just start accessing 100k+ Utah customers. Ripe for disruption.

I know nothing about your story, but based on what you said it sounds like it’s optimized already and the line-use fees+admin overhead pretty much explains why no one can offer better pricing. Again I know only what you wrote here, but that’s my uninformed opinion.

I’m not sure. The ISPs are charging $150+ for 10G plus you pay Utopia network $30/mo for the fiber line. So $180/mo total for the cheapest ISP.

Sure, but the ISP has to pay for transit, too. You aren’t going to be able to get settlement free peering as an ISP like that.

I don’t have a ton of knowledge in this area and am probably missing some things. But, I figured if they’re already able to offer 10G for $40/mo when they have to run their own physical infra on top of the peering and data center costs then they should be able to do it at a higher margin without. That higher margin could be a boon for them towards building their own infra in their current service area as well.

If they kept the same price then the total cost to the end user would be ~$70 for 10G. Which is less than half of the cheapest option available.

Transit costs aren't always the same in all locations

That’s nice!! $40/mo? That does not sound like US price.

It is. Sonic in northern California. More places soon, but scaling out physical infrastructure build has its own learning curve and often larger lead times. :)

Just got hooked up in Oakland and loving it! Paying $70 less per month vs Comcast

I was in Dogtown a year ago, but Sonic was just reselling AT&T to our address :(

Still very worthwhile to not have to talk to a Telco, but I had to rent + use AT&Ts garbage router because it did some proprietary negotiation step (With the ONT? There was a blog post where someone reverse engineered what was happening).

Now I'm in a smaller city and Sonic claims they have plans for us some day. For the time being, it's the Comcast nightmare.

Edit: and AT&T fiber was asymmetrical. Disappoint.

I'm in the same boat. I work there and can only get the resold AT&T service because my neighborhood was built pre-wired for only Comcast and AT&T, making it hard to service.

Still, like you noted, not having to deal with AT&T is a great selling point to a lot of people, and we're known for good support, so I'd like to think we're providing as good a service as those people can get until we build out our network in their area.

We have aggressive plans for the future, so I hope we're in your area sooner than later. :)

Same. About six months ago I started see Sonic trucks all over the neighborhood, five or six at a time covering multiple blocks. I asked if they were installing Fiber and was told no. Not much later, a sales rep knocked on my door. I said yes to the service before he could even greet me. It's been great.

Please come to the Peninsula (with fiber, Millbrae). Tired of my only options being Comcast or VDSL at max 20mbps :/

I'm not sure the exact areas we're building and planning, but it's possible we're coming soon. As I understand it we hit most of the low hanging fruit in the bay area already (above ground infra, aka poles), so we're trying to use advancements in trenching and hitting some slightly less dense areas than we previously targeted to be able to serve new areas at the price point we've set.

Dang! Good for you guys!! I’m in TX and ATT+Frontier have a firm grab on the infrastructure here. Charter is trying but still cannot compete with symmetric fiber.

I'm not sure, but I get the feeling that independent ISPs are having a bit of a resurgence, so maybe you'll be lucky and someone will look to serve your area. Or you can try it yourself. :)

I imagine it's easier now than it used to be to find some areas with good beauty and above ground infra (poles) that have space, and make a business plan and point at others that are doing it successfully as justification.

Or maybe we'll be there in a few years. At the rate we want to expand it's not impossible. :)

There's a 10 Gbit cable going in on my street for $60/month in Florida. Some interesting things happening in internet infrastructure these days.

I would love for this to happen on my street in Florida...

I think it's way too rural at this point however, given that I am technically in unincorporated area.

What street? Are schools any good?

What isp are you referencing?

I can get 6 Mbps for $40 here. Good stuff.

Lol and I’m sorry. I’d almost rather go with starlink at that point.

That's what I pay for 10Gb in the US. I think I'm a customer of the ISP the GP is referring to.

Getting 10G symmetrical May 1. I won't be able to use it all, though. Using a router that has a 10G port, but the stuff it does with packets means there's a limit I've heard of around 3G, but we'll see. Speed test is the first thing I'll be done once it's installed.

EDIT: USA, CA here. sonic.net. $40/mo with 3 free months. They'll pay up to $200 termination charges. Not affiliated with them.

My understanding is that cat-5e, for short runs (45m) you'd find in a house, can do 10gig. I'm still running 6a for all my new runs, but probably didn't need to.

And there’s a good chance that 404mm’s cat5 would pass as cat5e.

On point #1, you can probably do 10Gbps (assuming it's actually Cat5e), as long as it's under 50m or so. If it's just a run inside your house, it's likely well within range.

Yep, you beat me to this comment. I ran cat 5e through my house in 2009 and didn't bother with cat 6 because no cable runs would be that long. I assumed there would be 10Gb consumer switches by now, but I haven't seen any yet.

There are consumer 10gbit switches now, though they are on the pricy side at around $400. I bought 3 of them for our datacenter and 2 of them broke within 2 years, so I definitely regret buying consumer grade for our operations but professional gear was ridiculously expensive then. Upgraded to ubiquity now, hopefully it's more reliable, though I guess it's on the prosumer side if you ask your average net admin.

Well, when I say consumer, I mean unmanaged and less than $100 for 4-8 ports. I assumed that when gigabit over UTP was formalized in 1999, I wouldn't have to wait a quarter century to upgrade, but here we are. And to be fair, gigabit still basically does what I need.

The power usage on 10GB copper is nuts so most of the cheapest switches come with SPF cages.

I'm guessing it's less than the one or two dozen 60 watt bulbs I used to use to light my house. But yes, I've heard this for a long time - at my work we used twinax or fiber often for top of rack switching and that's one of the reasons I was told. Not sure if twinax is much less power intensive, but I do have a buddy who's using regular old coax in his house (moca or something like that) as a multi-gig backbone (still less than 10G) now and he likes it. Nowadays that I have more money, I would have run fiber alongside the cat 5/6 cable to future proof things, but I'm told terminating fiber is a technical challenge. I guess there's no free or even cheap lunch for 10G yet.

What I've heard is that 10Gb over copper struggles with heat and power consumption, though it is technically possible.

Yes, it’s 5E. I thought that cat5e maxes out at 2.5GBit. Honestly, none of my devices would benefit from 2.5Gbit in a significantly way. I run some cloud backups in the night and it’s ok if it takes extra 5 minutes.

You can try, depending on length 10Gbit are achievable.

I've been able to get 10G over a shoddy 100ft cat5e run. Give it a shot! It's fun!

Nope, it does not work at 10 Gbps; just 5Gbps for short runs (30m) and 2.5Gbps for 100m. I never tried for very short runs (1-5m), but it does not work for 50m at 10 Gbps.

> Back to original question and with genuine curiosity - what is 25Gbit for???

From a switch perspective, it makes more sense to carve out 4x 25G ports vs 4x 10G ports if you have 100G switches. In a single rack unit you can fit ~32 100G ports, which can then be broken out to 128x 25G ports. That’s more port density than a 1U 48 port 25 switch/line card.

I live in New Zealand, 1Gb uncapped fibre is available in most urban areas at reasonable cost (sub $100/month) - I keep getting feelers from the fibre company to upgrade (to 2/4Gb) and I honestly can't imagine why I'd bother - I work at home (have for 20 years) but I seldom find myself waiting for the net, we don't have a problem when everyone's streaming .... and I'd have to upgrade ALL the household infrastructure - routers/wiring .... and we'd still be limited by wifi speeds (which is what everyone outside my office uses).

I used to design 2D graphics accelerators, one thing we learned was that perceived performance is not linear, in reality there was a performance region where people would always crave more, and another where things were fast enough and people were happy but wouldn't open their wallets for massive increases .... at that point 10x seemed more like 1.5x) - I suspect that something similar applies to home net performance too

The only purpose of 25 Gbit/s at home that I can genuinely see is "because I can".

Going above 1 Gbit/s can have (very limited) practical sense because you can hit that with e.g. Steam downloads, changing a 5 minute wait into a 3 minute wait. You can also basically stop caring about QoS on your uplink once you're at 1 Gbit/s with one human user, or anywhere past 1 Gbps with more than one human user.

For running cables, seriously consider fiber. The hardware isn't as expensive as it used to be, and it solves a lot of problems: You can (based on common sense, not sure of your building code) stick fiber in power conduits since its non conductive, you don't have to worry about potential differences etc., it's thinner (easier to hide/more wife-compatible) and once in place, you will be able to use the same fiber for higher speeds just by swapping the SFP's at the ends.

Yeah, Steam takes much longer "preallocating space" than actually downloading everything for me, so a faster line does not help much there.

>$80 for symmetrical 1Gbit fiber and have

Init7:

1/1 Gbit/sec CHF 64.75

25/25 Gbit/sec CHF 64.75

It's the same price if available.

Have you tried >1GbE Over your cat5? I’ve done 10GbE over short runs (100-150ft), and 5GbE or 2.5 should work over longer runs…

Wifi6E is around the corner for >1GbE wifi.

I have a feeling that infrastructural upgrades like these will be instrumental in taking spatial computing and entertainment to the cloud. Currently VR applications live in your headset / computer, but no reason why they have to. Of course, that's not its immediate use, rather a use case that might be triggered / enabled due to this move.

In my opinion, 5gbit would be an useful upgrade (probably not worth that amount of money). In practice 1gbit is normally the limit of downloading in my experience. However, with multiple devices doing updates, downloads, streams actually having 5gbit and limiting each device to 1gbit ensures a high speed all time. Especially with roommates/family.

I recently put some fiber (AOC, to be precise) in a multilevel European building - brick and concrete. It was not trivial, but not a PITA.

25Gbps is for multiple transfers from multiple sites to multiple computers. For example I transfer large files for work (up to 100 GB) from multiple sources and it takes a long time and consumes all the bandwidth in the building. If I download at lower priority (implement some QoS) then it takes even longer. I can see 25 Gbps eliminate the bottleneck and allow everyone in the building to have their regular use (video conferences, movie streaming, gaming, whatever) without me impacting. I would never use more than 1% of that bandwidth on average, but I would peak more than 1 Gbps quite often. Yes, 10 Gbps would be more than enough, but if the cost is the same then 25 Gbps seems more future-proof, at a cost (SFP28 equipment is much more expensive than SFP+).

25 users of 1Gbps service? Obviously, a simplification, but something along those lines is how they tend to be advertised

My initial grumpy reaction was the same. Then again a quarter century ago I probably would have thought: “Would I ever really need more than 56k? Takes only a few minutes to download all of Les Misérables during which I could read maybe 4 or 5 pages.”

Answering for myself, backups. My SO and I both WFH. We generate a lot of data, personally and professionally. Most of that backs up to a local server, which then backs up to two off-sites. That server also backs itself up (plex, etc.).

I have to schedule it all to run at night because it will saturate the network during the day otherwise.

I have 1Gb/40Mb @ $100 internet for reference.

Cat5 can be capable of multi-gigabit over runs less than a couple hundred feet. Don’t count yourself out yet.

> Back to original question and with genuine curiosity - what is 25Gbit for???

What a pointless question. I wish people would stop worrying about this instead of simply making it possible, so we can figure out what to do with it.

… but that’s what I’m asking. What can I do with it? As I mentioned above, I have 1Gig and can go up to 5Gig. What can I do with it??

Maybe you can't, but somebody will find a use for it. Everybody should have access to it and I'm tired of the excuses made for not improving this infrastructure "but nobody needs it!!!!".

I have friends who do video editing, work from home. They're throwing around 300 GiB files regularly and for them 5Gbit would be a godsend.

You can start a company, as a private person, and expect everything to work when you run it from home. Just as if you were to run it from anywhere else.

I suspect the metaverse will be a bandwith hog. This matters where you centralize or distribute the world model; ditto for the rendering.

What about 8K, 16K, etc video streaming? Not much content today but I expect them to become mainstream eventually.

... Are you serious?

We got 1Gbit in 1999-2000 at work. Upgraded to this crazy IBM standard for wall sockets. Turned out to be unnecessary. We had 10Gbit at work in 2006. When I switched to one of the 'big internet companies' in 2008, every fucking server was supposed to have 10Gbit. In 2013, every fucking server was supposed to have 40Gbit.

We're now in 2022. 8 years later. And people argue about 25Gbit at home.

Ḯ'm a bit ashamed.