The Internet's Physical Backbone Is Maxing Out

CommanderFrank

Cat Can't Scratch It
Joined
May 9, 2000
Messages
75,399
Forget about running out of IPv4 addresses, we are running out of backbone. No, not that kind of backbone, the phat pipe type of backbone and we are pretty much maxing out how much Internet traffic can be sent through the existing infrastructure.

If you up the power beyond a certain point, the fibers becomes light-saturated and the signal degrades. And we’re very close to reaching that capacity limit, according to researchers who convened at a Royal Society meeting in London this week to discuss the matter.
 
That is kinda misleading since the backbone of the internet is very different from country to country. I some places in europe it is still copper and people dig it up to sell metal to scrap metal companies. In the US the backbone is an eight foot across pipe that holds oc48 lines around the quote unqoute dark fiber which is the fiber optic lines that go between cities along the major highways. That is why only certain people are allowed to work on federal high ways... the backbone in the United Kingdom where this was looked into could simply be using less cables... Because just like copper where we simply added more cables to a T1 line to get an OC48 like with fiber we can simply add more glass strands... Personally I think it is silly that we use visible like since even shifting to ultraviolet light from visible red light which they use as a simple way of being sure the fiber connection is on, increases the amount of information they can fit between time intervals. a bit that is on is peak and bit that is off is a valley.
 
As long as they can ignore my menial traffic flow when they consider what to start blocking :p
 
In working for a backbone provider in Canada I call the article to be very misleading.

When we build out stuff we all of our links we designed to allow us to constantly up the wavelength of the fiber i.e. the speed of it. Between the two major cities in our province we run a giant bundle of 40Gb/s links (like 240) that is shared with all other providers. When we hit a certain threshold we will start upgrading that to 100Gb/s.

We also have have backup rings (we run 17 completed rings at a minimum of 10Gb/s bundles) that can take the load at a minor increase in latency. All it takes is proper design and good equipment - the issue is wireline stuff isn't "sexy" and most wireless based companies won't spend the money.
 
In working for a backbone provider in Canada I call the article to be very misleading.

When we build out stuff we all of our links we designed to allow us to constantly up the wavelength of the fiber i.e. the speed of it.
...
Appreciate your post.
To be clear though, increasing the speed is reducing the wavelength.
 
To fit more waves in a time period, the wavelength must be shorter.
higher frequency = shorter wavelength
 
You can fit more than one wavelength on a fiber pair.

Even with cheap metro DWDM equipment you can fit 80 channels of 10gig on a single pair of fiber.
 
To fit more waves in a time period, the wavelength must be shorter.
higher frequency = shorter wavelength

So you are saying....if I have a system capable of 80 wavelengths I.E. Cienna or something along those lines. When I go from 10G to 40G to 100G I won't be able to keep 80 wavelengths?
 
So you are saying....if I have a system capable of 80 wavelengths I.E. Cienna or something along those lines. When I go from 10G to 40G to 100G I won't be able to keep 80 wavelengths?

He's talking time based multiplexing and not wave division multiplexing.

And the last time I looked depending on hardware a 100gig was actually 4 waves bonded together.
 
you can say what you like.
If you increase the frequency, the wavelength gets shorter.
One is inversely proportional to the other.
 
He's talking time based multiplexing and not wave division multiplexing.

And the last time I looked depending on hardware a 100gig was actually 4 waves bonded together.

Not on DWDM. 80 channels is 80 channels more or less, whether its 10, 40, 100, or 200 on those wavelengths.


As far as 4 waves bonded together you are thinking of 100GE client. The two standards were 4x25GE(which is the dominant standard now) or the 10x10GE standard which didn't work for shit.
 
Not on DWDM. 80 channels is 80 channels more or less, whether its 10, 40, 100, or 200 on those wavelengths.


As far as 4 waves bonded together you are thinking of 100GE client. The two standards were 4x25GE(which is the dominant standard now) or the 10x10GE standard which didn't work for shit.

Sure if you can afford coherent you can get a 100gig wave but good lord have you seen the pricing on those optics!? At any rate, all metro targeted DWDM gear still uses 4 channels/waves, at least all gear we've seen and spec'd out over the last year from MRV, ciena, ekinops, etc.
 
i was expecting the removal of net neutrality to be pointed out as the solution. (no, i'm not in favor of twisted buzzwords like "internet fast lanes")
 
i was expecting the removal of net neutrality to be pointed out as the solution. (no, i'm not in favor of twisted buzzwords like "internet fast lanes")

Fast lanes weren't a solution to the problematic bandwidth we sometimes encounter. That was give priority to certain traffic for when our BW did choke, we still would of had the same issues.

Article is miss leading but still fun to speculate. I remember in the late 90's when we were getting smashed with all sorts of BW issues as broadband was getting online and all the ISPs and all the major backbone providers were crumbling (that was sarcasm).

I'm sure now with 1gig internet connections and unlimited BW slowly being introduced we will see more articles like this.
 
The big problem we are really going to start seeing is Router Vendors just can't keep up with the mega growth. Imagine that amount of 10GE bundles which max out at around 16 members or so.
 
A lot of the vendors have 96 channel 100G wave systems. You can mux 10G links over those 100G waves for 9.6T of throughput per fiber pair. Also some of the higher end systems are allowing 32 or 64 port LAGs. Plus you can leverage ECMP to gain multiple paths.

The per Mb cost of transit keeps falling so its not being sold like commodity that is running out.
 
In working for a backbone provider in Canada I call the article to be very misleading.

When we build out stuff we all of our links we designed to allow us to constantly up the wavelength of the fiber i.e. the speed of it. Between the two major cities in our province we run a giant bundle of 40Gb/s links (like 240) that is shared with all other providers. When we hit a certain threshold we will start upgrading that to 100Gb/s.

We also have have backup rings (we run 17 completed rings at a minimum of 10Gb/s bundles) that can take the load at a minor increase in latency. All it takes is proper design and good equipment - the issue is wireline stuff isn't "sexy" and most wireless based companies won't spend the money.

Just wanted to add that regardless of the state of backbone, 99% of canadians have access to 10/1 high speed, the state of highspeed in canada is a disaster. Unless you live in some rare area in downtown toronto/montreal etc.
 
Reminds me of chip fabrication...every time the doomsayers prophesy that "we're going to hit a wall and crash and burn" some company or other has proved them wrong...;) Every single time. Sometimes people get hung up on thinking that there's only one way to skin a cat...:D People really need to be leery of doom & gloom prophecies these days--behind every one of them is a slimy snake-oil salesman with a costly "cure" for the particular doom he's dreamed up (this week.)
 
It has been a long day :) Just got my wisdom teeth out on Saturday one should not post smart stuff (and/or work on hard things lol) when getting off general
 
https://www.axia.com/plans

We are starting to offer Fiber to people homes in rural communities through Alberta Canada. We don't match Google Fiber's pricing (well I don't think) but we also have no data caps and/or restrictions/throttling either. The large companies don't touch the small markets and it makes me kinda sad that I could get 100/100Mbps straight into a proper non-congested backbone in a tiny town.
 
A lot of the vendors have 96 channel 100G wave systems. You can mux 10G links over those 100G waves for 9.6T of throughput per fiber pair. Also some of the higher end systems are allowing 32 or 64 port LAGs. Plus you can leverage ECMP to gain multiple paths.

The per Mb cost of transit keeps falling so its not being sold like commodity that is running out.

In theory yes you can do 32 or 64 group lags....but in reality not so much. Those vendors are having a lot of errors in reporting and snmp etc etc. I mean if you don't care about knowing when your dropping packets or other telemetry.....
 
“You can’t get an infinite amount of capacity in a fibre,” Andrew Ellis at Aston University in Birmingham, UK, told New Scientist.

This is a "UK" problem not a US one. Gues what? you can lay down more fibre!!
 
Back
Top