Slashdot Mirror
Do 'Bandwidth Bullies' Abuse Their Positions?
Zannah writes "An article on Forbes' website talks about the 'bullies of the Internet': the few providers who dominate much of the internet backbone, and who charge -- rightly or wrongly -- smaller organizations for access and peering to the backbone. It raises the question og whether the Net really *is* as open and distributed as touted, or whether it's in fact dominated by monopolies. As someone who works for one of the mentioned 'bullies' (UUNET), I think it's a bit one-dimensional, but certainly worth reading." This articles raises some interesting points, especially for rural customers and ISPs. Of course, in much of the world it's far worse anyhow.
When I was doing some reading about the backbone of the internet a couple years ago I ran into this same thing. Basically it's not that there are bullies on the backbones but there are two theories of backboning. The first camp thinks that getting rid of their packets as fast as possible gets them the most benifit. The second gets the packets as close to the destination as possible then dumps them. You can see how if everybody did one or the other it would work well, but the backbones are mix and match. The camps that want to get it as close as possible want many peering points and the 'dump the packets' camp wants just a few. Incidently these peer points are extremely overloaded at this point.
It has always been that you can buy special routing priveledges on these backbones, so I don't see the real point of this article.
On a side note, there is also other companies setting up 'internets' where they will garentee bandwidth, latency, and peer points to other networks, with complete security. I can't remember any names off hand, but there is a whole new demension starting to open up.
The separate question is whether IHighway is big enough to be a peer or if they've got to buy transit service. Most of the Tier I providers will peer with other large nationwide carriers, but will treat local ISPs like they treat other non-ISP businesses, and charge money for service unless there's a really interesting traffic mix.
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
>jerks who get their fancy new cablemodem and then
/dev/null.
>download 800 gigabytes of porn in one evening,
>starving off the entire feed for everyone else on
>that part of the network - or the xDSL guys who
>do the same thing.
I *PAY* for for 640K/640K bandwidth on my DSL line. And short of illegial activities such as spamming or DOSing, what I do with the dedicated bandwidth I *PAY* for is my own bloody business. Even if all I do is send 640Kbps of random noise back and forth to a friend who *PAYS* for the same rate and dump it all to
Now, if you want broadband on the cheap and want to go with a shared service like a cable modem, that's your own affair. You get to save some $$$, but, in exchange for that savings, you take your chances of being on a shared net. Got a problem with your performance? Reread your service agreement and see if you have a valid complaint. And, if so, take it up with your provider.
As for myself, I *PAY* for 640Kbps, both ways, no limit. And if I want to use the bandwidth I *PAY* for, that's MY own business, and I'm both legally and morally in the clear. If an ISP can't deliver the bandwidth it guarantees, then they SHOULD NOT SELL THAT BANDWIDTH!!!
john
Imagine all the people...
I've been thinking recently that the States, particularly those without a lot of high tech industry, should lay a bunch of fiber between the population centers and take bids on who wants to run it. A winning bidder would have to guarantee peerage with all the big boys (UUNet, SPRINT, ATT, etc. etc...).
This would sure give IntraState eCommerce a big boost and make the State a more attractive place to build an eBusiness.
These days, you are more likely to be closer to a business in San Francisco than a business in a city 20 miles away in terms of Internet topology. If the states did something like this, it might help the local economies.
With good connectivity, you could run your eBusiness hub practically anywhere. Heck, if getting top technical people to relocate to West Virginia is a problem, you could leave the people in San Francisco and move the computers, phones, call center drones, etc. to wherever you can get the cheapest real-estate, power, phone lines, drones etc. and remotely administer/develop from Silicon Valley.
You can't get cheap real-estate or ANYTHING in Silicon Valley, that's for sure.
Seems like building a hospitable eBusiness infrastructure by laying some fiber is a cheap way for the States to help the business/jobs climate in that state. All this available bandwidth tends to help your schools, libraries, and people too.
There are SOME things that governments do well, like Dams, roads, etc. It's a chicken and egg thing. You can't get SPRINT to lay lots of fiber between Charleston, and Parkersburg, WV until there's demand, there'll be no demand as long as the connections are too slow to support eBusiness.
I'm not too worried about the States laying lots of taxes and regulation on the segments that they paid for. After all, if people find one State too oppressive, they can always go to another state where the environment is more friendly.
-Jordan Henderson
It's true. .it just can't be done.
It's also amazing how many don't realize that, given the current network architecture, there isn't enough core bandwidth to give everyone even 1Mbps clean..
At least your provider is up-front about what your services are. I'd rather they say 'here is what you get for $x/month, and here is what it costs if you go over' rather than 'You have been busing your connection. Please refrain from abusing it by running servers or we will discontinue your service'.
We're in the wrong environment. They paid for the early internet highways, but these days the social benefit principles that built the old highways are a little scarce in the legislative branch. They're building internet 2 stuff, but it's not really for public use. Any future projects of this type by the government would involve unacceptable amounts of monitoring in the name of national security and protecting IP, so I think it may be better if we leave it in the hands of companies who will cut corners (read: not pay for monitoring) to save a few bucks.
Incidentally, grandparents may really be a font of wisdom after all: mine says the old highways were a giant jobs project in the name of rapid movement of defense equipment, note that all major thoroughfares and offramps are large/strong enough for tanks to drive on at full speed, certain three-lane strips are extra wide and flat for use as emergency airfields, etc. The suburban sprawl part was what the automobile industry would call a "happy accident." I thought all he knew was baseball.
-jpowers
-jpowers
A T-1 line around here costs about $500/mo on average with something like a $2,000 install fee. That is only for a point to point line in the same area code and exchange.
My question is how much does it actually cost to provide this line? Somehow I don't think it comes close to that.
Let's take a look at other so called telephone "services."
I could easily go on about how these scumbags abuse everyone, but I'm sure that I'm just preaching to the choir. Why should these people be allowed to gouge everyone with their monopoly? The fact is there is little cost difference between analog and digital services. Most of them use the same copper that analog phone service does so why the big price difference?
I knew someone was going to try to sound superior by bringing up history.
Well now, we wouldn't want to be doomed to repeat it would we?
Look, we all agree that government involvment in the initial protocols, but was ARPANET popular? No. The internet only grew AFTER it was released into the 'public domain' as it were.
Just because A follows B, it doesn't mean that A caused B. I would argue that the Internet grew fastest after the development of the World Wide Web and GUI web browsers like Mosaic and Netscape made it easy for most people who were intimidated by Usenet, Gopher, Telnet, and the like to use the Internet. Government was largely irrelevant in either case; it neither helped nor hurt the growth of the Internet.
Is corporate control better? Of course not, but one could argue that there is no corporate control over the net. There is only government control at the request of corporations (and who would turn down a request from your contributor?)
One can make that argument but realistically, there is some level of corporate control over much of the Internet from the big corporate-owned backbones, down to the corporate owned websites that flow through their pipes.
Even Slashdot has corporate ownership these days -- granted, Rob and Jeff had a good lawyer who ironed out good contract that guarantees their independence, but I think you see my point. Large, bureaucratic organizations like government and large corporations will always seek to control. It's often desirable to have them acting as a check on each other rather then letting one of them run hog wild and do whatever it wants.
Does this
You know, I've heard this spin before, and I feel the need to debunk it. What the carriers are doing is exactly what they should be doing with the information they have. As you put it, What the big guys do is dump your traffic off at the local public exchange point at the soonest opportunity. Which is exactly what they should do. If the packet is destined for one of their customers, they should get it there directly. If the packet is not destined for one of their customers, they should get it off of their network ASAP. Why? Because the carriers have no idea what each others network looks like. Not at the BGP level. All any of them know is what AS path they should use. In other words, they don't know how many hops away a destination is. They just know which carriers the packet needs to go thru. So they get it to the proper carrier as quickly as they can. Would you rather that UUNet bounced your packets around for a few extra hops instead?
To make this more clear, let's make up an example. Say an MCI customer in Washington DC is sending a packet to a Sprint customer in San Jose. Let's also stipulate that MCI and Sprint peer in DC and SJ. Someone needs to carry that packet across the country. Who should that be? The answer is Sprint, not MCI. We can complain that MCI is dump[ing] traffic at the soonest opportunity, and we would be correct. Because, MCI has no way to tell that the destination is in California. Only Sprint knows that. So, MCI gives the packet to Sprint ASAP (in DC) because the alternative is for MCI to bounce the packet around it's own network at random. Remember, only Sprint knows where Sprint customers are. Is it "fair" that Sprint has to carry the packet, and that MCI just dumped it as fast as it could? Maybe yes, maybe no. It doesn't matter. The way backbone routing works mandates that it be done this way. Don't like it? Feel free to try and improve on BGP.
My company had DSL service via Concentric. In the six months I worked there and the service was available, we had one outage, lasting over 12 hours (!).
In our new location, we have a spiffy new UUNET T1. So far, it's been great. The people are good, the service is super-fast, I'm delighted. Of course the service has only been on for three days, but so far, I'd say it's worth the money.
A company you've never heard of will give you a T1 for around $ 999 a month. UUNET's T1 was $ 1,300. Considering that the line is a pretty hefty percentage of our business, I thought the price premium was reasonable. It used to be much higher (like 200%-300% over what the cheapest vendor would charge). Note that the complaining former UUNET customer was in the UK, where telecommunications costs are much higher, which distorts the picture.
I thought the article was muddled and confusing. The problem is that it costs a lot more money to drag cables to remote areas, so naturally direct T3 service is not available there, at least not at a price you might want to pay. Even with peering arrangements, you have to pay big fees for space in the cages and access to the other players.
I would certainly get a UUNET T1 to my house if and when I could afford it.
D
----
this post brings up a very interesting point, IMHO :)
the internet is simply a collection of networks.. I can create my own internet, you can create your internet (commonly reffered to as an intranet, these days).
what's my point? We chose to let them be our backbone providers, it made sense at the time.
We also chose to have our IP addresses assigned by a central authority
The internet is not owned by me, its not owned by you, not by anyone. What they offer is a high speed connection - but it would be very easy for us to tell them to get lost. How?
Its a wonderful tool called ipchains
SSL Certificate
Most peering (even among the big boys) is based on traffic equity. If you are pushing more than twice the traffic you are pulling from a peer you will probably have to pay a "settlement" fee, based on the imbalance. All backbone operators
try very hard to maintain a balanced traffic profile with their peers.
The long-distance backbone market has been getting more and more monopolized for a while, and some people have tried to draw attention to it. But not many people noticed, because it's a sector that's not obvious to a layman.
Don't forget, these companies didn't "build it all themselves" like they claim, and are not innocently trying to get a return on their investment. Some, like phone and cable companies, were granted their monopolies and rights-of-way by our government, supposedly to use them for the public good. It's like broadcasters "owning" a broadcast frequency. They now claim to own what was originally public property by its very nature, either tangible or intangible.
What is it they are whining about? That sprint gets UUNET connections for free? Or that UUNEt gets sprint connections for free? Or that they don't? He's mad becaues he has to pay if he wants high-speed access where it doesn't exist already? He's mad because Sprint won't pay for it?
Well... the market *IS* free, and anyone *IS* free to start their own network, lay their own fiber, and sell their own bandwidth. If the big boys want to peer with each other, what's the problem? They could always just bill each other, but it would probably just cancel itself out, and they would end up spending more in paperwork than anything else...
What? no bandwidth is available in Buckwheat, Kentucky? Oh.. perhaps, then, there is MONEY TO BE MADE if someone were to finance a big, fat pipe back to civilization! Or perhaps not.. and if not, tha'ts probably why it isn't there in the first place.
Exactly why they charge, sometimes though to nose, for ACCESS to those 'prohibitivly expensive' lines. They paid for them, and are charging according to the cost to maintain and provide those lines. But the point is, that you are by no means FORCED to use these supposed 'bullies' lines. Several alternatives exist, all which cost MORE then they do..
-- I'm the root of all that's evil, but you can call me cookie..
Having a look at their web site this page mentions that "* The P-NAP, and its ASsimilator technology are proprietary and patent-pending " - so we really don't want these guys to be the future of the internet. No matter how good their technology is, if they are moving such basic internet standards to proprietry forms, they are not a company I feel people should support.
tangent - art and creation are a higher purpose
postmoderncore - art and creation are a higher purpose
I subscribe to xDSL, and pay a lot more for it. If I want to use that connection to download an entire Linux distro in one shot while playing EQ on another system, then that's my business. If my ISP could not handle the traffic, then they would not sell me the connection.
I'm sure you would like things the way they were "back in the day", when we all were using the Internet for nothing but Gopher and Newsgroups, and nobody transfered anything at speeds above 9600 baud, but that day has passed. Try to get over it.
Information wants to be anthropomorphized.
The rail analogy can be taken further. Rail companies regularily share traffic over their lines. The in most cases, empty cars are sent back to their origin in the exact reverse order they travelled to reach the destination.
This results in empty cars travelling less than optimal routes back to their origin. The purpose of all this is to ensure that the companies who profitted from moving the loaded car, pay for moving the empty.
I believe a similar business model is being applied here. The backbone providers share their loads more or less equally. The little guys have little to contribute in terms of large scale inter-connectivity so are out of the game. But a packet is a still a packet, it is never empty or full like a boxcar, all packets cost something to send and receive no matter how large or small the connectivity provider is.
This leads to an interesting possibility. The smaller providers could charge the Backbone companies for access to their customer base. The big providers won't find the smaller markets profitable, but as a group, the smaller providers would represent a good piece of traffic. Podunk may not count for much, but 10,000 Podunks might make a Chicago.
Yes, but that's not what they're doing. If you look at the matrix of UK peering agreements, you'll notice that UUNet doesn't peer with quite a lot of the other Linx members. All Linx members are backbone ISPs, not small local outfits, but UUNet (and some of the other larger players) still refuse to peer. And that's refuse to peer, not "we'll peer with you if you pay us". Because at the end of the day, refusing to peer with the smaller guys protects their monopoly, and that's far more valuable to them than the income they'd get by selling peering agreements.
"The invisible and the non-existent look very much alike." -- Delos B. McKown
And just like the airlines can not blame the passengers when they all show up, you can't blame somebody who is paying for unlimited broadband for actually using it.
Information wants to be anthropomorphized.
I'm not saying there is no problem. Clearly there is. But whining about things that clearly are available is not going to solve the issue.
BBN was an old, old company when they got involved in the Internet. They had been an accoustics company before hand.
The problem is that these companies aren't playing by their own rules. UUNet's rule is in order to peer with them you have to have four geographically diverse POPs connected together with a backbone of at least 45Mbps (DS3).
I was hired as a consultant by a company trying to peer with UUNet. They had three POPs (SF, Chi, and NYC) and were looking to build a fourth all connected with T3's. We started talking to UUnet and they said they were no longer establishing any new peering arangements.
They did offer to connect to us via a DS3 for only $50K per month.
The Economics of Website Security
Albuquerque is remote but it's not that remote. New Mexico is poor and as a result we have suffered from poor connectivity. That's no one's fault but it is shortsighted.
;-).
I'm not aware of ATT currently offering connections to anyone locally. The pop you are referring to must be brand-new or not yet active.
That said, I work for a company (osogrande.net) that does have UUnet and Qwest connectivity. You don't have to resort to the Verios of the world just to get Internet connectivity in Albuquerque. You also don't have to pay $10K extra per month for a DS3 (you just have to negotiate well
As an aside, John Brown moved back here from California about 3 years ago, so he does know what he's missing.
It's really amazing how far the Internet has come without settlements between the backbone carriers. It started without settlements because it came from a DOD/NSF background, and continued that way because nobody could agree on settlements should work.
I looked over InterNAP's web site, and I fail to see how they are doing anything special. It appears that they simply want to run the large access points. However, changing the legal structure of the access points doesn't imply any real change from a technical standpoint.
What's really needed are simply more peering points. If InterNAP can help throw a few more into the mix, that's great. But I don't think they are doing anything special from networking point of view. When you get down to it, peering points are difficult to get started. It costs a backbone provider real resources in terms of cost, equipment, and time to drop a connection into a peering point. Unless it's already a popular point, why bother? I mean, who wants to be the first one into a peering point, right?
It's also important to keep in mind that there are a lot more peering points than just the MAE's. Some are big, some are small. Some are popular, others are pretty much failures. Sometimes big backbone companies peer with each other multiple times all over the country. Other times, there are egregious failures of companies to peer. My favorite example is USWest and Qwest, which are in the process of merging and both have HQ's literally within blocks of each other in Devner, but don't peer!
So while peering can definitely use some improvement, I don't think it's all that bad. In fact, I would hazard a random guess that most slowdowns are due to the smaller ISP and web sites on the edge of the net. A lot of ISP's grossly oversubscribe to make ends meet. A lot of smaller web sites can't afford the bandwidth they really need if they get popular or a flash-crowd hits. Actually, a lot of days I'm amazed that this thing still works at all.
That pisses you off, eh?
Know what pisses me off? Someone selling me a 2Mbps connection, and then trying to make me out like some kind of asshole for actually *using* it.
That, and people who tell me that 'internet' access = 'web' access.....
We already have a USENET 'law' that establishes the first person to bring up Hitler in a conversation. Now do we have to have one for Ayn Rand?!
Pope
Freedom is Slavery! Ignorance is Strength! Monopolies offer Choice!
It doesn't mean much now, it's built for the future.
The bottom line is cost. (I work in the NOC of a very large ISP)
Since a T1 costs 40 times as much, ISP's place that much more priority on issues involving a leased line than a telco line.
Since DSL lines are usually put in place by the local telco, they are considered a "consumer" service, and most likely get an SLA guaranteeing service in a matter of days, or weeks!
With a T1, most SLA's have service guarantees in a matter of hours. The ISP takes a great deal of responsibility for the circuit, with the option for many corporate services that are not (yet) offered with DSL.
Maybe in the future, once the DSL demand is reigned in by local telcos, the technology will replace more costly T1/E1 service. But until it does, the T1 will remain the best and most stable choice of businesses.
It raises the question og whether
OG not like pay backbone company. OG raise question.
OG get Invalid Form Key. Og try opening other browser window with IE5 Win98. Og hit submit now. Og hit Cowboy Neal with club if not work this time.
For all intensive purposes, "whom" is no longer a word. That begs the question, "who cares"?
...think of web services as trains, bandwidth as rail; everyone wants to start a new train company-- great discounts, beautiful dining cars... all the perks. No one wants to lay new rail, and they all complain about having to pay to run on the extant rail. Yeah, it takes capital, but it's the price of freedom.
Do something about world hunger. Click here
Oligopolies can be illegal if they start colluding to maintain inflated prices, for example (the lysine market comes to mind). But rarely is a structural remedy imposed in such a case.
"If one is really a superior person, the fact is likely to leak out without too much assistance" -- John Andrew Holmes
If I'm not mistaken, the companies that provide the so-called back bone to the internet have to pay hard earned money to pay for all the fiber, staff, land, etc. that they need to make it happen. Unless they start getting these things for free, let's not jump to the easy conclusion that they are "bad guys".
-- You see, there would be these conclusions that you could jump to
When I was pricing a T1 for my home I checked out uunet and the other carriers. Granted, I don't live too far from a hub, but UUnet's pricing wasn't too far above everyone elses (about $350 more than other local carriers once I got done negotiating).
Plus, they were the only company who guaranteed latencies in north america and transatlantic, as well as constant server monitoring and free paging for detected outages. That and a free days credit for 1 minute of downtime, whether on their end or the local loop.
I couldn't afford the price then, but when I can, they'll be the guys I go with. Yes they're more expensive. But you do get what you pay for.
Of course I use Microsoft. Setting up a stable unix network is no challenge
this got me thinking - why doesn't the government start building public "roads" for the internet with tax money? They've been doing it since cars were invented, in real life, so it should be a natural progression, right?
Just think what the world would be like if major companies owned all the main highways and freeways?
Hell, I don't think this'll actually happen anytime soon, but it's a thought.
when the rain comes, they run and hide their heads. they might as well be dead.
If I got this striaght.. UUNet. Sprint.. AT&T and the likes allow traffic from each other to freely flow over each others networks.. which seems to be qui pro quo... You let some of my traffic use your network and I'll let some of your traffic go over mine..
But the problem as Forbes sees it is that an ISP complains because they want access to this bandwidth so they can make money. but don't want to pay the access fees ??
I guess I don't follow..
The bigger guys seem to be trading services, but the little guys have nothing to trade (besides $$$) so the providers have to make them pay $$
I appreciate the insightful commentary for once...
...
However, I would like to point out something: I'm not saying that anyone is doing something they shouldn't, I am saying that the public peering model is fundamentally broken, and that what we currently have results in asymetric routing and high packet loss at peak times. Why should traffic hit these bottlenecks if it isn't necessary? Why not use private peerings with multiple backbones? Extra hops are introduced, and eficciency is lost. That's why I think what they are doing over at InterNAP is so cool: they want to re-architect the system so that all participants receive maximum benefit. I'm not a nutcase, and I know I keep bringing them into this discussion, but if you ever hear their CTO talk, you'll get chills at the ideas they work with....
These large backbone providers are essentially advertising (Qwest, for example) "We have the largest, fastest backbone in the world!", but who cares? If you get dumped at the competitors network immediately, you spend all your time in transit as a freeloader on someone else's network. BB prov's don't pay each other for this transit, so it gets lower priority on everyone else's network. That's not a good system, because participants have competing agendas.
Check my Go-related blog for beginners: DGD
Nothing wrong with the spirit of the posting but if this is going to be rated as insightful, someone should get the facts right.
Take note of the fact that UUNet and Sprint are both owned by MCI. MCI is still negotiating with the government antitrusters as to whether they get to keep sprint plus some of their other aquistions. Many tears in the corporate boardroom.
So where is your competition when they all get bought out by Time-AOL-MCI-ATT-TweetyBird-... ?
When you are dancing with wolves, never limp
If the provider advertised truthfully, then:
a) Customers could use all the bandwidth provided to them in any (legal) manner they see fit, and:
b) The provider would take technological steps to limit customer bandwidth usage to an amount that won't have a detrimental effect on others.
Unfortunately, providers are selling more bandwidth than they have capacity for. It's a heavy temptation, when you have tons of customers who do light surfing for an hour or two a day. You can make a lot more money by selling them bandwidth that they will probably never use. Of course, sooner or later, someone actually does use it, at which point the provider backpedals and bends its terms of service to somehow restrict this person's usage (which he/she is actually paying for).
Personally, I'd rather know I'm paying for limited service and leave it at that. Sure I'd like huge bandwidth. But I'd rather have less bandwidth with a straightforward policy, than "unlimited" with strings attached.
Best regards,
SEAL
Now hold on a second, there.
There's a very underreported problem with the way that peering is set up right now on the network. "Peering", as it has been, is only going to become worse and worse, because of the kludges we have had to provide to keep things just as they are aren't enough anymore. 90% of all routed traffic going from one backbone to another goes through one of the MAE exchanges, or in other words, through a public access node that is paid for with tax dollars alone. This being the case, those facilities are massively insufficient for the load they bear, which is why on peak times, you can get 20-30% packet loss. This is absolutely unaceptable. And the worse of it is, the more loss, the more traffic, as the TCP protocol simply has hosts re-send what didn't get through...
Some people in the high up networking world are getting a clue, and have actually solved the problem. For more information, a better and more thourough explanation, and more technical details on the solution, go to this site .
To continue, the tier1 providers refuse to peer with smaller ones, as the article suggests. UUNET gets pretty much whatever it wants because they own 30% of all the destination ip-space. What the big guys do is dump your traffic off at the local public exchange points at the soonest opportunity, making their competitors carry the load. But of course, since they all do this, everyone gets a worse QOS unless they just happen to be going to someone located on their own backbone.
Thats not good, and that is what I think this article is (rightfully) pointing out.
Check my Go-related blog for beginners: DGD
We have to do something about this. This is the first article I have seen in years relating to what I have been griping about. It's a great article, but what is going to happen about it?
I believe we need to form together and get things changed or create another nationwide network. Get a bunch of us to form a utility company and employ thousands across the US, we'll use our won resources and create a new free network.
...and I'm not sure we should trust this Kyle Sagan either.
The Tier 1 backbone ISPs are today competing in a free market. No price regulation, and no monopolies. So the price is what the market will bear. "Peering" is done for mutual self interest. UUNET needs good access to Sprint's subscribers, and Sprint needs access to UUNET's, so they peer for free. That's how the handful of Tier 1s work -- it's in their mutual self interest.
Now you have little guys who have no content to speak of. So it means little to UUNET or Genuity to have good access to them. Yet they want free peering just like the big boys! Hey, I wish my house had free peering over a T3 too, but somebody has to pay for the Tier 1s! They <B>sell</B> backbone service to smaller ISPs. If it were free, where would the money come from? Repeat: This is a free market. The price is not set by the government, but by haggling until both sides' mutual self-interest is met. Not the way some Americans might be used to it, but the real world is full of bargaining. BTW, smaller backbone ISPs (Tier 2s, who almost by definition are the ones whose web sites call themselves "Tier 1", don't always pay a fixed price for Tier 1 upstreaming. They negotiate based on mutual self-interest. As they get bigger, they acquire leverage.)
Ditto to the smaller country players who whine about paying for bandwidth to US peering points. If they had content that the US providers though it was more vital to have, then they might split the cost.
If I were running an ISP, I would frankly be a lot more worried about the chance of Roadrunner Cable (and the various DSL providers) shutting me out of the broadband market than the possibility that my parent connection is gouging me for a few extra bucks.
Information wants to be anthropomorphized.
The Public Exchanges? MAE-East and MAE-West (metropolitan Area Exchange) handle 70-80% of all traffic that goes from one backbone to another., and are paid for by the government. It's the one place where everyone can peer for free, so it is where the smaller backbone providers peer at, so they don't have to pay exhorbitant rates to UUNET or the other fatties...
InterNAP is getting it right, and is beginning to get noticed around here and elsewhere. Check out their web page for a good explanation. They provide to some of the biggest blue-chips in the country, including M$ and Amazon. They have a different way of doing this stuff that solves the public peering problem.
Check my Go-related blog for beginners: DGD
If my ISP could not handle the traffic, then they would not sell me the connection.
... sure they would. Internet economics works sort of like airplane economics --- airplanes regularly oversell seats on certain routes based on the presumption that [x%] won't show, and sometimes they're wrong, and have to bump people.
Uh
ISPs do something similar: they regularly sell ~120% of their available bandwidth under the assumption that *it won't all be used at the same time*, and then once in a while there is a problem with it.
The biggest problem is with cable modem providers, where bandwidth is explicitly shared between multiple subscribers, but it exists with other types of ISPs, as well.
I'm sure you would like things the way they were "back in the day", when we all were using the Internet for nothing but Gopher and Newsgroups, and nobody transfered anything at speeds above 9600 baud, but that day has passed. Try to get over it.
The problem isn't that people are downloading more than they were back in the days of gopher, but that the rate at which the total available bandwidth has increased is *lower* than the rate at which the total amount of data being routed has increased.
The major issue with this that I see is when do you reach the point where you're peering rather than purchasing bandwith? For example: You start a small ISP. You buy a T1 or a T3 from a backbone provider. As you grow, you add a few other connections and multi-home your network. At what point do you become big enough that those pipes to the backbone providers become peering points, and should be free? If peering was free for anyone, once you got big enough you, you'd get all of your bandwith for free? Even if you only apply this rule to providers with a backbone, what constitues a backbone? What if you have two datacenters across town from each other? That could be considered your "backbone". The way I see it is that providers shouldn't have to give free bandwith unless they recieve something in return. That return could just be faster access for their customers to reach sites hosted by the opposing company. Once an ISP gets big enough, other ISPs should have no issue at all wanting to do free peering with them, which in most cases is how it works at this point (in most cases). Gah. Ok, I'm done babbling.
Oligopolies do screw the consumer (Gasoline, Automobiles, Backbone providers); however, the exsistance of oligopoly is not illegal.
As long as they aren't in collusion with one another, in which case it is explicitly illegal.
(Note that the DOJ is currently involved in an antitrust suit against Visa and Mastercard).
goes through the MAEs or NAPs is bogus. If somebody's
paying you to deliver traffic for them, there's no way you'd want to deliver through the heavily overloaded public peering points if you've got a working private connection to the destination carrier, because packet loss is too high, except for overflow traffic and disaster recovery. The reason you build private peering connections is precisely to avoid the NAPs, as well as to save costs and occasionally deal with geography.
That doesn't mean that the peering isn't in the same cities as the NAPs - The Tier 1 providers have a lot of bandwidth there, so a packet may occasionally go more distance to get to a peering point, but the connections are still private, so the load management depends only on the two carriers involved, not the entire group of users on the NAPs. I don't know UUNET's peering objectives, but AT&T's objective is to deliver at least 90% of our off-net traffic using private connections, and only use the NAPs when there isn't a better way to get somewhere.
(I'm using the term "peering" to refer to the technical aspects of the connection, not whether money changes hands in the process - that's going to be driven by traffic balances, special applications, and whatever ego-based negotiations happen between contract lawyers :-)
Hosting centers are clearly changing the balance of traffic on the net - Exodus and the non-carrier based hosting centers connect to several major ISPs, and most of the major ISPs either run hosting centers or are partnering with them, and they're being built as fast as the real estate mavens can get them built. (Who'd have guessed that Bay Area building contractors would be one of the industry's critical resources? :-)
Also, the emerging caching business may skew traffic levels differently between NAPs and private peering - I'm not sure which direction, and some of the skew is affected by different caching architectures - some caching systems put big shared caches on the outgoing side of a multiple peering point as opposed to more but smaller non-shared caches on the incoming side.
Ryandev's assertion that UUNET owns "30% of all the destination ip-space" isn't very clear - they certainly don't own 30% of the IP address space, though their market share for direct-connection IP customers is probably about that big, depending on which pieces they get to keep in the merger (e.g. they sold of much of MCI's business to Cable&Wireless during the Worldcom-MCI acquisition.)
Also, the MAEs and NAPs don't just run on public money - they charge for connecting to them, and I'm not sure if the NSF subsidy is still there. The important issue is that they're a finite solution to a near-infinite demand, and traffic can grow faster than they can scale. While lots of traffic has moved off to private peering, they're still a huge and critical resource, so they both need to use bleeding-edge technology and cause real havoc if the technology doesn't work well enough - and there's nowhere else to test it adequately before using it....
Internap does have some interesting BGP path-manipulation technology; I've seen it described but haven't played with it directly. Their business model is to buy transit from the big fiber-based ISP carriers and sell services to end users, so their success depends largely on how well they can manage and implement their networks, as well as on their BGP technology. Unlike the Fortune columnist whining about how mean and nasty UUNET is and how Somebody Should Do Something, they're going out and building competition, and good luck to them. (Hey, it's the Internet - they're our competition and our customer also :-) And no, they don't have a POP in Albuquerque either - their closest is Denver.
(Disclaimer: I work for AT&T, though I primarily do Layer 2 and Layer 1 technology rather than Layer 3, so I may be out of date on some details. This is my personal opinion; if you wanted an official position, you'd be reading a press release or talking to a sales person, not reading a rant on slashdot... :-)
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
You can. All you need is two 24/7 servers to handle your dns and you can handle your domain's traffic. All you need to do is find somewhere to hook up to the internet.
That, of course is the issue. The "best" place to hook up is through one of the major backbones. If you don't care about performance you can hook up through anyone, (and they hook up through one of the backbones...).
Then just get your ip blocks from internic and you're good to go.
Of course I use Microsoft. Setting up a stable unix network is no challenge
It's not dominated by monoplolies, for one reason. Anyone, *ANYONE* can setup a highspeed connection between two access points. It doesn't make sense to say that these guys are bullies. You can buy this kind of access from *anyone* who wishes to offer it.
I could, right now, buy a fiber line from one side of the country, and start piping data from here to there. That's it. That's *OPEN*. And heck, people could start routing their IP packets over my line, no effort on their part.
This is just a story for the sake of writing a story. There is no basis behind this. PERIOD.
-- I'm the root of all that's evil, but you can call me cookie..
Second, the Internet became commercialized when BBN (Bolt, Beranek and Newman) was created (if I remember correctly) specifically to sell NSFnet to. BBN was the first commercial backbone. Everybody else got blocks assigned by IANA and either built their own or leased from BBN. Nobody "chose" that format, it just grew that way.
I remember a few years ago the screaming when the big boys announced that they would begin refusing to peer with smaller networks. It makes sense if you think about it -- you have to draw a line somewhere. There comes a point below which you aren't big enough for big backbones like UUNet or AT&T to care if they lose all your traffic.
The flip side of this is, big sites like eBay or Yahoo have leverage because everybody wants access to them. There's opportunity for them to leverage their popularity into lower or no connection fees if they're savvy enough to manipulate it.
And really big sites like AOL are now backbones in their own right, to the point where magazines like BoardWatch rate AOL's network as a backbone provider.
-- Old Man Kensey