Fibre versus Starlink Discussion

[Lindamon]

Low latency is not what you think it is.

Latency is the time it takes to get a response. Say you are playing a streaming game, latency affects how long it takes the game to register your button presses and for the action you commanded to appear on your screen. Similarly if you are using remote desktop and you type sometimes latency affects how long it takes for the result of your typing to appear on screen.

By the way, fibre is much better than an ancient T1 line in this respect. As well as offer 20Gbps of bandwidth on consumer lines, ping times are extremely low as well. If you are using high bandwidth interactive services that's what you need, both of those things.



Was the cost of installing phone lines or electricity lines recouped? Or the cost of building roads all the way out there that hardly anyone uses? What about water supply?

Maybe not, but it was worth doing anyway.



Yes, many places are short sighted and dumb.

I know next to nothing about the details of internet service types, but I can tell you you that water supplies are not generally run to rural areas. In rural areas the vast majority of water is supplied by wells dug on the property. Sewage is handled with holding tanks and drain fields, also on the property. Up until 19 years ago the 2 mile long road to our neighborhood from the state road was dirt, our neighborhood had to pay to have it paved. Electricity supply and landline phone service costs are certainly recouped, FPL and Ma Bell see to that.

 

[bkp_duke]

Any cable provider will have higher latency than a fiber provider.

The transition from Coax to Fiber at the junction box is usually 12-15 ms.

 

[banned-66611]

Also these servers don't exactly prioritize ICMP so variations are not indicative of anything much.

 

[banned-66611]

Engage my brain? Says the guy that cannot defend his arguments.

MOST fiber is trenched, not up on poles. Why? Because of tornadoes, hurricanes, morons driving cars into poles, etc. you don't want the risk of a fiber break of crap strung on poles. MOST fiber in Japan is actually underground, NOT on poles, you nit wit.

I run a business that GUARANTEES 5-9s (99.999%) uptime to different datacenters on 3 continents. It takes HOURS to repair fiber (it's a bitch to splice), so no one in their right mind runs fiber on poles except to residences, where you don't have uptime guarantees.

Why don't YOU engage YOUR brain and do some research before you spout out your personal, uneducated opinions as FACT.

Most fibre in Japan is on poles. The poles are unbundled, lots of utilities share them.

 

[banned-66611]

Canadian standards for this rests with the Provinces. All Provinces but Nova Scotia insist that fibre cables must be buried. We have been lobbying our Provincial Government to use overhead cabling, as most of the rural areas (like me) have their hydro delivered by utility poles. Their reply, so far: Get lost.

So rather than use the polls the solution is it send up tens of thousands of satellites and offer an inferior service?

 

[banned-66611]

You completely ignored my statement on latency. Yes, lower latency is "better", but I would not trade the 26ms latency of the cable HFC for the 16ms latency of the dual T1 line, as the bandwidth available results in major congestion and very high latency when in use during the day.

Cox 100/20 Small Biz Cable (Tucson AZ):
ping -c4 8.8.8.8
PING 8.8.8.8 (8.8.8.8) 56(84) bytes of data.
64 bytes from 8.8.8.8: icmp_seq=1 ttl=117 time=26.6 ms
64 bytes from 8.8.8.8: icmp_seq=2 ttl=117 time=28.4 ms
64 bytes from 8.8.8.8: icmp_seq=3 ttl=117 time=24.9 ms
64 bytes from 8.8.8.8: icmp_seq=4 ttl=117 time=28.0 ms

Level 3 dual bonded T1 (yes, 3/3 service, no users on at the time of the ping) (Tucson AZ):
ping -c4 8.8.8.8
PING 8.8.8.8 (8.8.8.8) 56(84) bytes of data.
64 bytes from 8.8.8.8: icmp_seq=1 ttl=115 time=17.9 ms
64 bytes from 8.8.8.8: icmp_seq=2 ttl=115 time=16.4 ms
64 bytes from 8.8.8.8: icmp_seq=3 ttl=115 time=16.5 ms
64 bytes from 8.8.8.8: icmp_seq=4 ttl=115 time=16.3 ms

I have been involved in internet engineering for the last ~25 years, built proxy server environments, made Internet Access usable in K12 environments where it previously was not usable due to bandwidth constraints. I have deployed plenty of Point to Point wireless links due locations being unable to get effective bandwidth.

I have also been involved in enterprise networking for a considerable amount of time.

You keep missing this is NOT a one size fits all, and that there is plenty of room for StarLink.

I can even see eventually small towns will put in passive optical networks (GPON, etc), but use commercial StarLink for the uplink.

Just because we had no other choice but to dump huge amounts of $ into rural electrification and telephone (copper) networks, does not mean we should do that again with Fiber.

Plus many of those rural telephone deployments used microwave relay towers to get outside connectivity, not copper (coax or pairs) or Fiber.

Many small towns are still serviced with Geo Satellite or Microwave Relay towers for external connectivity to the PSTN, not Fiber.

We have tons of above ground pole mounted Fiber in Tucson AZ, yet the residential availability of Fiber is TINY. Still, anyone in Tucson AZ is not a primary StarLink customer, maybe out side of town in areas like Vail, but not in the city.

-Harry

Why do you keep posting about old T1 lines? Why but compare with something modern?

Your numbers look pretty terrible but that's not surprising with your equipment.

 

[hmcgregoraz]

Also these servers don't exactly prioritize ICMP so variations are not indicative of anything much.

Ok, so you want to change the goals again.

Yes, ICMP is not prioritized. The statement was that anything more than 1-2ms ping to Google was unacceptable.

All of the services, with the exception of the bonded dual T1 are very usable.

-Harry

 

[DMC-Orangeville]

So rather than use the polls the solution is it send up tens of thousands of satellites and offer an inferior service?

Oh yeah, piece of cake....
-Convince the Provincial Government, and their molasses administration to forego any studies, and immediately change legislation to allow fiber on poles
-Convince 1 of the big 3 telcos, or their smaller rivals, that it's good business to connect those bumpkin customers - they'll make money, we promise!
- nope that won't work with the telcos
- Instead, we'll lobby all levels of government, and convince them to shell out billions to the telcos, so that we can get fiber run out to us. Heck, the feds are giving everyone billions in the Covid-19 world, so they will easily be persuaded to slip a few billion out, and right away - like our $2000 CERB (sorry Americans, you'll have to Google that story). Also, throw in a few billion more, and we can bury the cable. The Provincial Government is off the hook!

Plus for only a couple of hundred billion more, and we'll run fiber up to the Yukon....and Northwest territories, and Nunavut. 100/10 to Pangnirtung!

Now, lets get this all done in say, 6 months! No problem, Yahoo!

Or

Get out of the dream world, and get a private industry expert who can cover the globe with a (slightly) lower quality of service....and they can spread their costs out amongst a MUCH broader base of customers, and sell it at a reasonable cost right away.
No cost to the tax payer.

I'll take the blue pill, Neo.

 

[hmcgregoraz]

Why do you keep posting about old T1 lines? Why but compare with something modern?

Your numbers look pretty terrible but that's not surprising with your equipment.

The equipment is mostly modern Ubiquiti and Mikrotik, and the latency has nothing to do with the equipment/CPE. All of the latency is upstream.

Most Tucson traffic routes to Phoenix via two separate fiber paths, one goes along I-10, along the rail road right of way, the other goes around the Catalina Mountains.

Most services such as CloudFlare and Google are connected to in LA area, and not in Phoenix, due to provider based decisions.

There is plenty of data center space in the Phoenix area, and I have worked with plenty of customers in those data centers.

The dual bonded T-1; I have been trying to get moved to HFC with Cox, and have had quotes lined up for months, but no progress from the business management, despite the cost savings. HFC has a standard install fee, but zero build fee. Fiber to that location was over a $10,000 build fee. HFC only recently became an option at the location, as a few years ago Cox wanted a five digit build fee to get HFC to the location.

As I have said, I have nothing against Fiber, but saying that StarLink is trash, means that you are completely out of touch with the realities of consumer internet and internet engineering.

Order of capabilities:

1. Fiber to the house (XGS-PON should be able to do 10000/10000 service)
   
2. Fiber to the distribution point (g.FAST, 100M = 900/? service, but less "shared" vs HFC)
   
3. Hybrid Fiber Coax (docsis 3.1 should be able to do 1300/200 service)
   
4. Fiber to the node (VDSL2, 250M = 300/100 service)
5. StarLink (100/50, probably growing to 1000/?? service)
6. ADSL2 (100/10 bonded is the best it gets)
   
7. WISP (depends on the deployment, could be better than StarLink if they have a good uplink)
8. 4G LTE / Fixed 5G (again, very dependent on the uplink)
9. GEO orbit satellites (ViaSat, HughesNet, etc)

We need a combination of ALL of the technologies above to get effective internet access to everyone.

-Harry

 

[scaesare]

No, scroll back and you will see I factored it in to my calculations.

The problem is your calculations and the thresholds you are suggesting are required are entirely arbitrary.

For the scrolling impaired...

For the reading impaired, please note I said help offset.... but nonetheless, continue.

consider that the satellites are at an altitude of 320km, so that's the minimum distance that the radio signal has to travel. Of course it's usually further because they are not directly overhead. And then it has to come back down too. So immediately you have at least an extra 640km to travel.

And that distance has added all of ~2ms to the overall latency.

2/3rds speed of light is a bit low for modern fibre but okay, let's go with that. Fibre can send a packet 422km in the time it takes Starlink just to go up and down, let alone sideways.

So the issue is with pulling data from a local CDN? Like streaming a video, downloading data, or loading a web page? You are expecting that additional 2ms latency is a showstopper?

Yet you say in a subsequent reply:

Latency is the time it takes to get a response. Say you are playing a streaming game, latency affects how long it takes the game to register your button presses and for the action you commanded to appear on your screen. Similarly if you are using remote desktop and you type sometimes latency affects how long it takes for the result of your typing to appear on screen.

Oh.. so now the issue is gaming. So the issue is either local CDN content, or the overall-latency induced by the parties connected to the game server (which isn't part of a local CDN).

So which is it? You can't have it both ways without moving some goalposts around.

Incidentally, do a search on acceptable gaming latency, and you'll see the ~30ms that Starlink is targeting is right in the sweet spot.

Halve it for the round trip. Most people live much closer than 210km to a CDN, so your response on fibre comes in before Starlink has even covered the distance to space and back.

As pointed out above, it matters not one whit for watching your Netflix movie. And the only real scenario requring low latency you've been harping on (gaming), isn't served out of a CDN.

Your "requirements" don't pan out.

 

[scaesare]

Any cable provider will have higher latency than a fiber provider.

The transition from Coax to Fiber at the junction box is usually 12-15 ms.

Hmmm... My initial FIOS installation was Coax from the ONT to their on-prem router (which i reconfigured to be bridge). I now have a twisted pair GigE connection straight from the ONT to my router/FW instead.

While I don't have records, I'm sure that earlier cable setup didn't add that kind of latency to the overall link, as I would get sub 10ms ping times to local speedtest servers.

 

[scaesare]

Why do you keep posting about old T1 lines? Why but compare with something modern?

Your numbers look pretty terrible but that's not surprising with your equipment.

You are missing the point that overall throughput is not only dependent on the application requirements, but a function of both latency and bandwith.

 

[banned-66611]

Ok, so you want to change the goals again.

Yes, ICMP is not prioritized. The statement was that anything more than 1-2ms ping to Google was unacceptable.

All of the services, with the exception of the bonded dual T1 are very usable.

-Harry

You don't understand at all. I'm saying your test methodology is imperfect, not that the goal has changed.

When did you say that more than 1-2ms ping to Google was unacceptable? What is your basis for that claim?

Can you explain why you have this affinity for the old T1 service? Why avoid comparing with modern fibre?

 

[banned-66611]

Oh yeah, piece of cake....
-Convince the Provincial Government, and their molasses administration to forego any studies, and immediately change legislation to allow fiber on poles
-Convince 1 of the big 3 telcos, or their smaller rivals, that it's good business to connect those bumpkin customers - they'll make money, we promise!
- nope that won't work with the telcos
- Instead, we'll lobby all levels of government, and convince them to shell out billions to the telcos, so that we can get fiber run out to us. Heck, the feds are giving everyone billions in the Covid-19 world, so they will easily be persuaded to slip a few billion out, and right away - like our $2000 CERB (sorry Americans, you'll have to Google that story). Also, throw in a few billion more, and we can bury the cable. The Provincial Government is off the hook!

Plus for only a couple of hundred billion more, and we'll run fiber up to the Yukon....and Northwest territories, and Nunavut. 100/10 to Pangnirtung!

Now, lets get this all done in say, 6 months! No problem, Yahoo!

Or

Get out of the dream world, and get a private industry expert who can cover the globe with a (slightly) lower quality of service....and they can spread their costs out amongst a MUCH broader base of customers, and sell it at a reasonable cost right away.
No cost to the tax payer.

I'll take the blue pill, Neo.

Sounds like the problem is that your government isn't really interested in universal service or the digital divide.

In Japan they deregulated the poles, i.e. the companies that own them are obliged to provide access to other companies on a non-discriminatory basis (same cost and conditions as they impose on themselves). That opened up the market dramatically.

There was also a universal service requirement. Having said that Japanese companies don't tend to be as short sighted as Western ones, they see fibre as a long term investment that may take decades to pay off sometimes. A lot like their high speed rail and many other things there.

The most bizarre thing about your argument is that all the same points could be made about SpaceX.

"Let's just launch 40,000 satellites, I'm sure there will be no problems with global spectrum allocation and the whole thing can be sorted out in six months! It will only cost a few billion, they will make money we promise! There won't be any environmental issues putting 40,000 objects in orbit and then continually replacing them every 5 years!"

 

[BioSehnsucht]

Hmmm... My initial FIOS installation was Coax from the ONT to their on-prem router (which i reconfigured to be bridge). I now have a twisted pair GigE connection straight from the ONT to my router/FW instead.

While I don't have records, I'm sure that earlier cable setup didn't add that kind of latency to the overall link, as I would get sub 10ms ping times to local speedtest servers.

He might have meant a hybrid coax/fiber setup not like FiOS where often it's coax from the ONT, but where it's fiber to the node/neighborhood, then coax for the last "mile" (not necessarily a mile) as employed by a lot of "cable" (vs "phone") companies

 

[banned-66611]

Most services such as CloudFlare and Google are connected to in LA area, and not in Phoenix, due to provider based decisions.

Phoenix to LA is about 600km, which is a bit less than the minimum 640km additional distance up and down for Starlink.

If you sit down and do the maths for a moment you will realize that fibre over ground is still going to be faster than Starlink. That is kind of missing the point though, obviously if you are WFH for a company nearer where you live the distance is much shorter, and as it becomes more of an issue more CDNs will arrive in Phoenix anyway.

The dual bonded T-1; I have been trying to get moved to HFC with Cox, and have had quotes lined up for months, but no progress from the usiness management, despite the cost savings.

It's sad that they are still installing old tech like that. Sounds like regulatory failure, in other countries they wouldn't be allowed to install it at all and would be required to roll out fibre at reasonable cost.

saying that StarLink is trash

I didn't say that, you are confused.

Fiber to the house (XGS-PON should be able to do 10000/10000 service)

Double that in fact, Nuro Hikari 20GS has been offering 20Gbps symmetric lines in some cities for a few months now.

 

[banned-66611]

And that distance has added all of ~2ms to the overall latency.

Yes, so why did you bring it up? You are the one posting all these dubious ping tests.

 

[banned-66611]

You are missing the point that overall throughput is not only dependent on the application requirements, but a function of both latency and bandwith.

That's exactly what I said. And I also noted that it's not just latency, but the variability of that latency.

 

[BioSehnsucht]

Phoenix to LA is about 600km, which is a bit less than the minimum 640km additional distance up and down for Starlink.

If you sit down and do the maths for a moment you will realize that fibre over ground is still going to be faster than Starlink. That is kind of missing the point though, obviously if you are WFH for a company nearer where you live the distance is much shorter, and as it becomes more of an issue more CDNs will arrive in Phoenix anyway.

You're assuming that fiber follows shortest path. Even ignoring possible switching delays, fiber rarely follows the shortest path, but the most economical at the time it was built out (which often means following cheap rights of way, like rail lines or along highways, etc).

i.e. https://www.reddit.com/r/MapPorn/comments/6bhovm/os_1432_787_internet_map_in_the_us/ (not a perfect representation, but close - someone pasted the raw map from http://conferences.sigcomm.org/sigcomm/2015/pdf/papers/p565.pdf on top of a map of the US to make it more understandable, though it looks like the attempt to scale the fiber map to fit the normal map was imperfect, it gives you an idea of the issue)

And I believe his point was that not only did traffic from Tuscon not route to Phoenix, but to LA instead of Phoenix, even though Phoenix is much closer (rather than merely be concerned with the Phoenix to LA route). So from Tuscon it's more like 800km along the shortest major fiber path to LA, even though the shortest distance is much less ignoring the actual fiber layout. Things get a lot worse for more rural parts of the country, too. Huge swaths of the country have no major fiber backbones or interconnection points near them, so many locations will be going well out of their way before they get anywhere useful.

If Starlink can use ISLs to get closer to the destination, or even just skip over silly provider routing decisions even without it (like Tuscon going to LA instead of Phoenix, can probably reach both cities from a single Starlink sat without ISLs and space-based routing, if they wanted to provide routing between them - slower than a direct path, but if a direct path doesn't exist ... though in this case the fiber exists, it just isn't where things end up being routed, so that's moot), then there's plenty of scenarios where Starlink might provide better latency than fiber simply due to the more direct nature (versus the often windy and arbitrary fiber paths), never mind Starlink providing merely good enough latency in general.

 

[scaesare]

Yes, so why did you bring it up? You are the one posting all these dubious ping tests.

1) I brought it up because you are touting the round-trip to the sat as if it were a significant disadvantage when you said, "So immediately you have at least an extra 640km to travel".

B ) To point out 2ms to the RTT isn't a big deal in the real world

III) You seem confused. Please go back and look at who's posting ping time tests (hint: it's not me). While you are at it, you may want to look at the points that poster is making with them. They are informative.