| Two problems, don't conflate them The Network World article is excellent; Johna (who is the head of Nemertes) is one of the smartest people writing in the trade press. But the press articles attacking her don't recognize two separate points being made.
One is total traffic volume. Not that it can't be provided, but it can't be provided at the price you guys want. The current growth rate of average usage isn't 100%/year, but full-rate (not Flash) video uses orders of magnitude more than web and similar classic applications. So it could become a problem easily.
The Internet is not one network -- that's why it's called Internet, a network of networks. So it does not go all crappy at once. Capacity is cheap in Manhattan, NY, not so cheap in Manhattan, IL, and even costlier in Manhattan, NV. That middle mile is a killer. I was talking today to a client whose WISP is the only near-broadband service in a town of 1000 people whose nearest town of over 50k is over 100 miles away, in Idaho someplace. We're looking at pulling FTTH. But middle mile capacity is incredibly expensive and limited. These are microwave routes to the nearest fiber. Backbone capacity in such places can't be treated as a zero-cost commodity.
Johnson's article also referred to a totally unrelated problem, IP backbone route proliferation in BGP. This is a scaling problem in IP itself which is leading to real "sky is falling" effects as the routing update mechanism will not be able to keep up. It makes being on the backbone costlier, too, regardless of how much or how little traffic you have.
IPv6 makes it worse, not better; it's no solution to anything meaningful. That's why some of us (who have formed the Pouzin Society, named after Louis Pouzin, as close as you can get to a real inventor of the Internet) are advocating the development of PNA as a new protocol stack. It is totally different, but still interworks with IPv4 (for transition) better than IPv6. Read up about it at »www.pouzinsociety.org/ . PNA decentralizes the model better than IP. It fixes routing-related issues like multihoming and mobility. It fixes the address issue too -- that turns out to be a red herring after all. It uses only names, not addresses, external to each (recursive) layer, so the address at any layer is local to a group of cooperating systems. But this is worth a different article, which I'll leave to the webmaster's discretion. |
|
iansltx
join:2007-02-19
Golden, CO
kudos:2 Reviews:
 ·Comcast
| The thing is, middle mile prices are falling. A year and a half ago Qwest came to my current WiSP and said "we can beat your current AT&T DS3 quote". They did, by $1000...so now my ISP is paying $9000 a month for a T3 with less IP space than AT&T gave them.
Now, I'm getting T3 quotes of around $4500. Still frightfully expensive since I'm out in the sticks (a T1 is $600 here if you ask in the right places, $1200 if you ask in the wrong ones). However things are looking up for rural places bandwidth wise.
The only people loud enough to cry about the bandwidth apocalypse are those who don't want to upgrade their (large) DOCSIS-based networks as far as I can tell. Oh, and wireless networks. And maybe VDSL networks (AT&T). Fiber optic providers aren't complaining, and DOCSIS 3 providers aren't either to any huge extent. Hmm... |
|
| In addition the biggest problem is that the middle mile is still mainly controlled by the same incumbents that control the last mile.
The incumbents are the problem at every part of the network because they are trying to maximize short term profits and will sacrifice anything to do it. |
|
| reply to fgoldstein
OK, I tried to to keep up, but you lost me there. 
Sooo... middle mile links are expensive in remote areas of Idaho = meltdown of the internet?
As far as internet backbone capacity goes, there are solutions to deliver speeds at the prices consumers want. But innovation like this will go on the back burner if billing by the byte is adopted.
32 Terabits per second on a single fiber
»www.att.com/gen/press-room?pid=4···id=26805
Cisco router released last year. Just imagine what they are working on now.
»newsroom.cisco.com/dlls/2008/ts_030408.html |
|
iansltx
join:2007-02-19
Golden, CO
kudos:2 | reply to Skippy25
Things really get dicey when there's no last mile as far as internet goes. You heard me right...Qwest and AT&T are the cheapest T1 providers for loop and port here. Verizon is the ILEC and they don't even have DSL. *groan* |
|
 espaethDigital PlumberPremium,MVM
join:2001-04-21
Minneapolis, MN
kudos:2
 ·Clear Wireless
| reply to 59126125
said by 59126125:As far as internet backbone capacity goes, there are solutions to deliver speeds at the prices consumers want. But innovation like this will go on the back burner if billing by the byte is adopted. The ironic part of that statement is that we have a plethora of backbone fiber today because backbone carriers bill based on usage. More infrastructure capacity = more revenue potential.
These are a bit like linking to things like GM's Hydrogen Fuel Cell Vehicles and declaring our dependence on fossil fuels solved.
There is a lot of existing infrastructure in place, and forklift upgrades are simply too costly for most companies to take on. The most cost-effective building block for backbone network expansion right now is 10GigE interfaces. There are currently efforts underway to finalize standards for 40GigE/100GigE, but the standard isn't slated to be completed until the middle of next year. One of the driving requirements for the 40/100 standard is that it needs to operate in existing 32/64 lambda DWDM systems, since carriers worldwide have billions of dollars invested in such systems.
This isn't a problem with backbone capacity, it's a problem with the costs associated with tapping into that capacity. There are a lot of discussions here about how a megabit is a megabit, and that's really not the case. The costs and complexity of upgrading a single backbone line from 1000mbps to 10,000mbps aren't the same as upgrading a hundred 10mbps connections to 100mbps. The bandwidth numbers are the same, but the interface costs / outage window negotiation / engineering effort are greater due to the increased scale. |
|
patcat88
join:2002-04-05
Jamaica, NY
kudos:1 | reply to fgoldstein
said by fgoldstein:The Internet is not one network -- that's why it's called Internet, a network of networks. So it does not go all crappy at once. Capacity is cheap in Manhattan, NY, not so cheap in Manhattan, IL, and even costlier in Manhattan, NV. That middle mile is a killer. I was talking today to a client whose WISP is the only near-broadband service in a town of 1000 people whose nearest town of over 50k is over 100 miles away, in Idaho someplace. We're looking at pulling FTTH. But middle mile capacity is incredibly expensive and limited. These are microwave routes to the nearest fiber. Backbone capacity in such places can't be treated as a zero-cost commodity. BS. The local CO has fiber or MW to the nearest POP guaranteed. Analog trunks area dead in the USA. Its just the local phone company laughs all the way to the bank with mileage charges for fiber that is almost fully depreciated in the books, and there are no other choices other than the ILEC (or MW) and their abusive mileage charges.
You would be surprised how much fiber there is in rural areas and along highways and railroads. Look for the "caution fiber optic" posts along any thick roads on the map, and look at the pole, I'm sure there is fiber there. |
|
patcat88
join:2002-04-05
Jamaica, NY
kudos:1 | reply to espaeth
Maybe its time for ISPs to invest in edge caching and for be cooperative with P2P efforts to put ASN prioritizing into P2P networks, split your upload traffic between 1/3 off-ASN, and 2/3rd same-ASN if you have enough same-ASN peers wanting data. No ISP should be complaining about same ASN traffic, the traffic costs are like one big LAN then. (P4P is sketchy in my opinion, ASN prioritization requires very little or no ISP participation, P4P sounds like your giving full peer selection ability and queue control to the ISP in real time.)
Remember in Asia with all those 100mbit and gigabit symmetrical FTTH, you can only do 100/1000 to other customers of your ISP, the speed to outside the ISP/outside the country looks like normal FTTH or normal cable modem service with speed in blocks of 5 and 10 mbitps, max 50 mbitps off the ISP/country. |
|
