Fiber Optics Lines Can Offer Much More

XEpsilon writes: "According to this article, the current usage of fiber optics lines is only .5% of the capacity fiber optics lines offer. The internet is also slowed down by old copper lines that are still being used -- converting from light impulses to electric signals is the major slow-down. A certain company, Cogent Communications, is offering unshared 100Mbps internet access for $1000, which is $500 less than the price of a T1." Interesting to note that Cogent bought just two strands of pre-existing inter-city fiber to re-sell the bandwidth. I'd easily pay $100 a month for far less than a tenth of the bandwidth they're promising -- let's hear it for economies of scale!

Comments

This is inevitable, but not the ultimate solution.

[spif]

The last line of the article sums it up nicely - this is all inevitable, not revolutionary or even particularly noteworthy. Cogent, Yipes et. al. are just taking advantage of a few relatively recent developments - the increased availability and affordability of dark fiber and Gigabit Ethernet hardware, and the maturation of DWDM. There's nothing terribly proprietary about that, aside from hundreds of millions of VC dollars and vendor financing plus the balls to build a very expensive network with a relatively small (albeit potential-customer-rich) footprint in the last mile.

Even if fiber deployment is on the way to solving the last mile problem (at least for business), it still doesn't solve the problem that really impacts the speed and reliability of internet service for most people: lack of fast, highly-distributed peering between networks. The speed of your backbone and tail circuits matters little if only a certain percentage of the entire internet is directly connected to them, unless you also have fast peering with other networks wherever possible. If there are dozens of "OC-192-or-greater-per-wavelength on DWDM on dark fiber" backbones, with Ethernet (regular, Fast and Gigabit) tails, each of which has only maybe a handful of OC-3 or even OC-12 peering connections to most of the others, the problem will still not be solved. If anything, it will get worse, because the users with faster last miles attached to faster backbones will expect proportionally faster service, and they won't get it.

There are people working to solve this problem, in various ways - running neutral peering facilities, aggressively seeking peering arrangements (although mostly in a few locations, unfortunately), buying lots of transit bandwidth from major providers (again, mostly in a few locations), etc. There are only a few who are truly solving the distribution aspect of the problem, though, by obtaining peering and transit connections to other networks in many locations evenly distributed around the country, in a mostly rational and consistent manner based on traffic analysis and other factors.

This is only one step in building a public internet infrastructure we can all depend on, but it is a crucial one.

fnord.

soon we'll have hundreds of metalab mirrors

[paled]

new Linux release out - time to grab an .iso:

100 Mbps / 8 bits/byte / 1024/1024 = 11.9 MB/sec

so for 50% utilization - figure 6 MB/sec

654 MB
------ = 109 seconds.
6 MB/sec

I'm going to need a faster CDROM burner.

Seriously - I still think that snail-mailing the latest Linux distro is a more efficient use of (very limited) resources than each person downloading their own .iso.

BUT - (Anne Marie) - its not up to me (or anyone else) to determine how people use THEIR Internet connection. The ISP can certainly cache content locally, thus only impacting other users on that part of the network when they access something that's bandwidth intensive. If the user is complying with the TERMS OF SERVICE of their user agreement, then it doesn't matter if they look at goat pr0n 24 x 7 x 365 - as long as it isn't kiddie goat pr0n.

Don't impose your morals on something that is Amoral - how people user their Internet connection.

this doesn't help me. =(

[fjordboy]

I have been saying this for a while, but people like me, who live out in the stix...aren't helped at all by these companies that are loading the internet with highbandwidth connections. Almost all of these new pipes and stuff are only available in large cities. I can't even get cable where I live, and the phonelines are old and not capable of dsl or anything. What would really impress me is a company that sets out to give country folk a connection to the internet. I get 36k on a good day....

Some people say that this will never happen because there isn't enough people living out here to make it economically viable, but I disagree. One satellite dish or something set on a mountain nearby could connect people within a 30 mile radius of me. That is a lot of people...and I know most of them would jump at the chance of getting high speed internet access. I would be willing to pay more than $100 a month for a good connection. Trout Run Pennsylvania needs broadband. :) Thanks for listening to my rant. erm..reading even.

Some corrections and additional info...

[trims]

After looking at all these replies, I thought I'd reply myself.

Some notes:

• For you British (and non-USA folks), you may not quite be aware of the size of the USA. Excluding Alaska and Hawaii, the population density of the USA is about 85 people/sq mile. For comparison, the U.K. (England, Scotland, Wales) has a density of 630.
• To most metropolitan areas (i.e. 100,000+ in the MSA according to the Census Bureau) in the US, there is at least one fiber trunk. Big places like the top 10 MSAs (NYC, LA, SF, Chicago) there are probably a dozen.

  If you live near one of these trunks, well, happy for you. However, remember than less than 50% of the US population lives inside one of the top 100 urban areas.

• Some cities (particularly those in the North) are replacing existing copper with fiber as part of the normal maintenance cycle. However, in much of the warm parts of the country, where the existing physical plant needs little maintenance, it's never replaced short of a big disaster ( Hurricane Andrew in Miami is a good example).
• Even hybrid Fiber/Copper layouts like the Cable Modem setups usually only bring fiber into the equivalent of the local CO - that is, fiber-to-the-neighborhood; the average distance from CO to house in a US city is well over a mile (remember, not as-the-crow-flies, but as-strung-through-the-pipes).

The real challenge is not the big urban areas - they can (and probably will) get fiber-to-the-curb within 10 years at the worst case. The challenge is getting it to the 30-40% of the USA that live in sub-100,000 urban areas, and even worse, the 10% or so that are classified rural.

It's the same problem the USA faced with electrification in the 1920s and 1930s. Only now, it's worse, since a far, far smaller percentage of the population lives in the harder-to-wire areas, there is even less of an incentive to fiber them up.

I can use my home town as an example: Meadville,PA, population 15,000. It's easily 40 miles to the nearest city which would have a fiber trunk into it, and while the town might eventually have fiber as part of the Cable Modem rollout, it's certainly not going to be fiber-to-the-curb, and the cable modem rollout will miss the other 70,000 people who live in the county (which is rather rural). Population density for my county: 82 people/sq mile.

-Erik

It's the "last mile" problem, stupid....

[trims]

(to paraphase the old "it's the economy, stupid" catchphrase of the last election...)

The article is plain wrong on several points, and makes naive and/or inane assumptions on several others.

First off, virtually all the long-distance communication wires in the US & Canada are fibre. Both voice and data travel over fiber (NOT copper) for about 99% of all inter-metropolitan traffic. Yet that fibre represents less than 10% (I forget the exact number, but it's less than 10%) of the entire physical communications plant in North America. The reason this is all fiber now is that it was by far the cheapest, easiest, and quickest return on investment portion of the network to upgrade. The article completely ignores the fact that the other 90+% of the network is extremely costly and time-consuming to replace with fiber and has a much, much, much longer ROI.

In places like Manhattan, where there is extremely high population density and communications demand, yes, fiber has been laid along (some) streets, and it is possible to put out a direct fiber lead to a large building that goes directly to a fiber line. If you're in on of the top 25 Metropolitan Areas in N.A., you might have a fiber line within a mile or so. Outside there, well, if you've got one within 10 miles, consider yourself lucky.

The problem still remains getting high-speed communications to places that don't astronomical population density. It's a hard problem, and condemning companies that provide more-or-less universal access for using copper is moronic. Cogent might be able to offer access to what, maybe 5% of the population? Compare this with WorldCom, the ILECs, MCI, and Earthlink, and all the other big ISPs, who can probably get between 95 and 99% of the entire population.

I also love the part where they seem to think that building their own long-distance backbone instead of using others is a panacea for network conjestion. Someone need to explain the concept of Peering and Wide-Area routing to these folks. Sure, it's nice if both the source and destination are plugged directly into your backbone, but the odds of this are what, virtually nill? If you want some real benefit from your own backbone, you have to connect directly with the majority of sites people want to access. In this case, you better beg Above.Net, Exodus, GlobalCenter, Genuity, et al. (all the big co-lo people) to allow you to run a line into all their co-lo sites. Oh, and since you're a small player, don't think that these folks aren't going to charge you for the privilege of hooking into their co-lo.

The other thing I find stupid is the implication that only Cogent has a "data-optimized" network. What a load of Marketing BS. If I'm running a packet-switched network, can I tell what I'm running on top of it (well, with ATM technically you can, but it's really all data)? Data? Voice-over-IP? Streaming Video? The article seems to confuse the concepts of packet-switched vs. circuit-switched with "data-optimized" vs. "voice-optimized".

Anyway, I'm sure others are going to point the myriad of crap in this article, so I'll stop here.

Cogent is offering a nice service, and has some interesting features that may point the way to how communications are done in the future. They're not really innovative in any sense, and I certainly wouldn't think of them as the greatest thing since sliced bread. It's an interesting company, and I wish them will. But the hype level is just a tad too high here.

-Erik

Moore's Law

[SETY]

There is no Moore's Law for fiber. I don't know where this guys source is, but the amount of data going down a fiber isn't going to double every 10 months. There are many new problems cropping up the faster you go. They just aren't easy to solve (basic physics).The polarization of the light as it travels down the fiber isn't understood.
And saying we are at .5% of the theoretical fiber capacity is great and all. The space shuttle also travels way less than the theoretical speed (c, the speed of light).
I'd put my money into optical parts being made cheaper. I wouldn't put my money on the amount of data through a fiber doubling every 10 months.