Re: OpenTransit (france telecom) depeers cogent
On Thu, Apr 14, 2005 at 11:28:00AM -0400, dgolding@gmu.edu wrote:
Richard,
Its easy to accuse people of criminality - that's what you just did. How about some proof?
Cogent's ratios are very very bad. That's why some people don't like peering with them. Being sent to sales is a common, if regretable tactic in depeering. However, dealing with Cogent on peering matters is incredibly unpleasant. I can understand networks and peering coordinators feeling that it just isn't worth it.
No, you got me wrong. I wasn't suggesting that anyone was doing anything criminal. I was suggesting that Cogent was now so large that if any of the folks who were pissed off at Cogent's disruptive pricing actually wanted to make an impact against them, they would need to get together and take action collectively. Teleglobe tried to depeer them, but ended up needing Savvis to transit the bits to maintain global reachability, which cost them a pretty penny while costing Cogent nothing. In the end, they promptly re-peered. I could certainly see anti-Cogent folks making the point that Cogent has amassed a customer base of people who are fully prepared for poor or partial connectivity, and who just don't care because of the price point. In the game of who can stick their fingers in their ears and hum the longest, I think Cogent will win. At least, if they continue to take on networks who don't like them one-at-a-time, and continue the policy of complete unreachability during depeering. :) -- Richard A Steenbergen <ras@e-gerbil.net> http://www.e-gerbil.net/ras GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)
On Thu, Apr 14, 2005 at 11:28:00AM -0400, dgolding@gmu.edu wrote:
Its easy to accuse people of criminality - that's what you just did. How about some proof?
Cogent's ratios are very very bad. That's why some people don't like peering with them. Being sent to sales is a common, if regretable tactic in depeering. However, dealing with Cogent on peering matters is incredibly unpleasant. I can understand networks and peering coordinators feeling that it just isn't worth it.
Ratio problems can be solved without hurting customers. Time and time again people use excuses - and most of them really are just excuses - to depeer people. Feel free, your network, your choice. However, as a customer of many big networks, we get to choose too. And we think we will stop choosing people who don't play nice with others. FT can claim "we saw their prefixes through Verio". Cogent can claim it was FT's "unilateral decision". Who cares? End of day, both networks are disconnected from a significant fraction of the Internet. The point of the INTERnet is INTERconnectivity. Not really that hard, people. -- TTFN, patrick
On Thu, Apr 14, 2005 at 11:28:00AM -0400, dgolding@gmu.edu wrote:
in depeering. However, dealing with Cogent on peering matters is incredibly unpleasant. I can understand networks and peering coordinators feeling that it just isn't worth it.
Just for the record, I've dealt with Cogent's peering people on behalf of a few networks over the last two years, and in my experience they've been extremely pleasant to work with. -Steve
On Thu, Apr 14, 2005 at 10:38:28AM -0700, Steve Gibbard wrote:
On Thu, Apr 14, 2005 at 11:28:00AM -0400, dgolding@gmu.edu wrote:
in depeering. However, dealing with Cogent on peering matters is incredibly unpleasant. I can understand networks and peering coordinators feeling that it just isn't worth it.
Just for the record, I've dealt with Cogent's peering people on behalf of a few networks over the last two years, and in my experience they've been extremely pleasant to work with.
Since I forgot to mention it in my last e-mail, Cogent does indeed have some exceptionally good folks working for them, particularly their peering people. However, we all know that in this particular case the issue is not about personalities, technical competence, ratios, etc. Anyone who doesn't know that Cogent has severely disrupted the IP transit industry's economy across the board with its pricing has been hiding under a rock for the past few years. They've recently slashed their prices to even more absurdly low new levels, and are actively targetting their peers' customers, particularly in Europe. Anyone who didn't expect to see exactly this kind of fallout as a result really hasn't been paying attention. What remains to be seen from all of this is who blinks first, if anyone else jumps in at the same time, and if any of it changes anything in the marketplace. -- Richard A Steenbergen <ras@e-gerbil.net> http://www.e-gerbil.net/ras GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)
They've recently slashed their prices to even more absurdly low new levels, and are actively targetting their peers' customers, particularly in Europe. Anyone who didn't expect to see exactly this kind of fallout as a result really hasn't been paying attention.
Well considering the market isn't growing as much as it used to the only way to grow is to take business away from other people. If OT can't deal with that on the street then they should close up shop and go sell kites or something. Clearly Cogent are doing something that gives them a cost advantage otherwise they wouldn't be able to sustain this, so rather than compete with Cogent, FT try to fiddle with Cogent's costs. If I was an FT customer and I'd seen this signal I'd be phoning Cogent now for a quote. If I was Cogent I'd be getting the champers out, for FT to take this action they clearly feel threatened!
What remains to be seen from all of this is who blinks first, if anyone else jumps in at the same time, and if any of it changes anything in the marketplace.
No, it won't change anything - other than for a short period of time customers will suffer, wouldn't it be nice if we remembered about the customers.
NOTE: Off-list, as I'm not sure this is on-topic enough to post. At 03:19 PM 4/14/2005, Neil J. McRae wrote:
They've recently slashed their prices to even more absurdly low new levels, and are actively targetting their peers' customers, particularly in Europe. Anyone who didn't expect to see exactly this kind of fallout as a result really hasn't been paying attention.
Well considering the market isn't growing as much as it used to the only way to grow is to take business away from other people. If OT can't deal with that on the street then they should close up shop and go sell kites or something.
This is an OK analysis as far as it goes. However, if Cogent has a business plan (and mind you I have not read theirs) that says "we will go borrow $10 billion, sell services at below cost until we bankrupt our competition, then jack up our prices" then that would not necessarily be OK. I don't know if Cogent is selling below cost. Amazon certainly built their business by doing so, losing money on every order. They did a nice job wiping out small book stores I expect. Are we better off for that? If Amazon were based in Korea and were selling products below cost into the US market, the US Government would scream "dumping" and insist on trade sanctions. The point of this is that competition is one thing, unfair competition by selling below cost (for more than a loss-leader or two) is not in the best interests of the economy or the future of any given industry.
Clearly Cogent are doing something that gives them a cost advantage otherwise they wouldn't be able to sustain this, so rather than compete with Cogent, FT try to fiddle with Cogent's costs. If I was an FT customer and I'd seen this signal I'd be phoning Cogent now for a quote.
If I was Cogent I'd be getting the champers out, for FT to take this action they clearly feel threatened!
What remains to be seen from all of this is who blinks first, if anyone else jumps in at the same time, and if any of it changes anything in the marketplace.
No, it won't change anything - other than for a short period of time customers will suffer, wouldn't it be nice if we remembered about the customers.
On Thu, 14 April 2005 20:19:41 +0100, Neil J. McRae wrote:
[..] Clearly Cogent are doing something that gives them a cost advantage otherwise they wouldn't be able to sustain this [..]
Do you seriously view it that way? See the financial analysis available on the web about Cogent and tell me the same thing again.
[..] If I was an FT customer and I'd seen this signal I'd be phoning Cogent now for a quote.
I want my packets to make it to the destination. For some Euros more I get real transit from real networks. See, all the world is fine again. He who wants to save even more money gets what he pays for. I bet your viewpoint would maybe change if you had nearly exclusively DSL traffic and hardly B2B- type traffic. I could be wrong, of course. Alexander
On Apr 14, 2005, at 3:47 PM, Alexander Koch wrote:
[..] If I was an FT customer and I'd seen this signal I'd be phoning Cogent now for a quote.
I want my packets to make it to the destination. For some Euros more I get real transit from real networks. See, all the world is fine again. He who wants to save even more money gets what he pays for.
I bet your viewpoint would maybe change if you had nearly exclusively DSL traffic and hardly B2B- type traffic. I could be wrong, of course.
"so-and-so said". We send more than enough bit over Cogent to both business and broadband users to be statistically significant, and the
You are wrong. I've heard the stories & rumors, but empirical evidence packets get there. "YMMV", but our _experience_ sending packets over AS174 is more than acceptable. Do you have any real experience sending traffic over / through AS174? To be clear, I could say the same about FT. Both run perfectly fine networks, and both give very decent pricing. This is not to say we have never had a problem - we've had problems with every network on the 'Net pretty much. But we were happy with both of them - until now. And if they don't get their collective acts together, I am certain they will both lose a lot of customers. -- TTFN, patrick
And if they don't get their collective acts together, I am certain they will both lose a lot of customers.
[Hijacking the thread here a bit] I think this is a good point you brought out. Neither provider is providing _full_ transit to their customers. If this becomes acceptable to a set of customers [say the multihomed ones] then providers will have to set a new price point probably significantly lower than the price point today for "partial transit". In this case, partial transit is defined by the provider as the set of ASes they may or may not have connectivity to at a certain point in the contract -- rather than the other case of partial transit where the customer decides what ASes are interesting. I think both providers in these depeering conflicts would be unhappy with where the customers would price this new partial-transit level of service at. Essentially, the responsibility of maintaining full-connectivity is pushed further onto those who have 2 connections.. now they need 3 or more to ensure they have 2 reliable/consistent paths to each destination of interest. I for one think that the providers that force their customers into this role should not be compensated at the same level as the providers that do a much better job of universal connectivity. In other words, "break routes at your own risk." Smart customers will be watching [sooner or later ;)].. disclaimer: I am not a customer of cogent, teleglobe, opentransit, and most companies that practice rabid peering discussions -- even for layer 1/2 transport if I can help it. Deepak Jain
Do you seriously view it that way? See the financial analysis available on the web about Cogent and tell me the same thing again.
Same could be said for many companies in our industry at the moment, I call them the zombies.
I want my packets to make it to the destination. For some Euros more I get real transit from real networks. See, all the world is fine again. He who wants to save even more money gets what he pays for.
I remember people saying that about Level3 and many others.
I bet your viewpoint would maybe change if you had nearly exclusively DSL traffic and hardly B2B- type traffic. I could be wrong, of course.
If I was solely in that business I would not be focused on building big international networks, its pointless and unless you take significant market share from the PTTs, notably in Europe (which btw is highly unlikely given that the various legislators and regulators fail on a near constant basis to level the playing field for all operators) - the chances of making a return are incredibly slim. Better to setup smaller national networks, exchange most of the local traffic at the local peering point and pay $transit_network for the other 30 odd percent, do a decent deal with $transit_network and you'll be quids in and have a much more managable cashflow cycle without the step upgrades to some chunky network that the majority of your customers don't care about. Neil.
On Thu, Apr 14, 2005 at 08:19:41PM +0100, Neil J. McRae wrote:
They've recently slashed their prices to even more absurdly low new levels, and are actively targetting their peers' customers, particularly in Europe. Anyone who didn't expect to see exactly this kind of fallout as a result really hasn't been paying attention.
Well considering the market isn't growing as much as it used to the only way to grow is to take business away from other people. If OT can't deal with that on the street then they should close up shop and go sell kites or something.
They should certainly do something about it any rate rather, instead of stick their heads in the sand and hope it all goes away. Depeering Cogent is doing something.
Clearly Cogent are doing something that gives them a cost advantage otherwise they wouldn't be able to sustain this, so rather than compete
Yes, its called losing money hand over fist. As long as people keep giving you more it makes it very easy to keep doing what you are doing.
with Cogent, FT try to fiddle with Cogent's costs. If I was an FT customer and I'd seen this signal I'd be phoning Cogent now for a quote.
You're missing the big picture. FT depeering Cogent increases costs for both sides, in fact probably more so for FT than Cogent. This has nothing to do with cost, and everything to do with FT believing that giving Cogent access to its customers for free is enabling Cogent to continue selling services below market rate, disrupting business for everyone. They may be right. Certainly if everyone took the same position, Cogent would be unable to continue to operate as it does now, so there is some truth to that belief. Despite what some people may think, peering is not a god given right that you are in any way entitled to receive. Two networks peer when they feel that there is mutual benefit, not before and not after. You may not agree with the other network's decision not to peer with you (and in many cases they me be wrong, and simply don't understand the benefits that are there), but it is their decision to make. If Cogent cared about connectivity to FT's customers, it would man up and pay for the transit to reach them. However, Cogent cares more about the long-term benefits of settlement free transit than it does about the short term benefits of being able to reach FT's customers today, so they choose not to pay for the transit in the hopes that FT will blink. Not that there is anything wrong with that, but that is the cold hard reality of the decision that they have made. It might even be the right decision. If you don't like it, no one is putting a gun to your head and making you buy Cogent (or FT) transit.
No, it won't change anything - other than for a short period of time customers will suffer, wouldn't it be nice if we remembered about the customers.
If you don't like it, buy from someone else. If you sell transit, now is an excellent time to pick up disgruntled customers from both sides. Either side could end the lack of connectivity if they wanted, yet they both clearly see business reasons not to do so. -- Richard A Steenbergen <ras@e-gerbil.net> http://www.e-gerbil.net/ras GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)
You're missing the big picture.
No I don't think so :-)
If Cogent cared about connectivity to FT's customers, it would man up and pay for the transit to reach them. However, Cogent cares more about the long-term benefits of settlement free transit than it does about the short term benefits of being able to reach FT's customers today, so they choose not to pay for the transit in the hopes that FT will blink. Not that there is anything wrong with that, but that is the cold hard reality of the decision that they have made. It might even be the right decision. If you don't like it, no one is putting a gun to your head and making you buy Cogent (or FT) transit.
Surely FT's customers pay for access to Cogents network and vice versa?
On Thu, Apr 14, 2005 at 09:58:08PM +0100, Neil J. McRae wrote:
If Cogent cared about connectivity to FT's customers, it would man up and pay for the transit to reach them. However, Cogent cares more about the long-term benefits of settlement free transit than it does about the short term benefits of being able to reach FT's customers today, so they choose not to pay for the transit in the hopes that FT will blink. Not that there is anything wrong with that, but that is the cold hard reality of the decision that they have made. It might even be the right decision. If you don't like it, no one is putting a gun to your head and making you buy Cogent (or FT) transit.
Surely FT's customers pay for access to Cogents network and vice versa?
In such a case, FT has done its part by paying Sprint for full transit service. It is Cogent who is not accepting the route from their transit, and who intentionally does not carry the global routing table. If I put up a filter on my transit that says I will not accept routes from you unless you peer with me, should your customers leave you because I did this? Doesn't sound very fair to me. I guess it depends how important I am, doesn't it? If I may, you sound like someone whom FT has depeered in the past? :) -- Richard A Steenbergen <ras@e-gerbil.net> http://www.e-gerbil.net/ras GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)
On Apr 14, 2005, at 5:16 PM, Richard A Steenbergen wrote:
Surely FT's customers pay for access to Cogents network and vice versa?
In such a case, FT has done its part by paying Sprint for full transit service. It is Cogent who is not accepting the route from their transit, and who intentionally does not carry the global routing table. If I put up a filter on my transit that says I will not accept routes from you unless you peer with me, should your customers leave you because I did this? Doesn't sound very fair to me. I guess it depends how important I am, doesn't it?
Is Cogent filtering the prefixes they get from Verio? Or is Verio filtering what they send to Cogent? Does it matter? I think you have a very good point - FT is buying full transit. Cogent is the one without full reachability. Doesn't mean that FT didn't know this would be a problem when they took the step, though. -- TTFN, patrick
On Thu, 14 Apr 2005, Patrick W Gilmore wrote:
On Apr 14, 2005, at 5:16 PM, Richard A Steenbergen wrote:
Surely FT's customers pay for access to Cogents network and vice versa?
In such a case, FT has done its part by paying Sprint for full transit service. It is Cogent who is not accepting the route from their transit, and who intentionally does not carry the global routing table. If I put up a filter on my transit that says I will not accept routes from you unless you peer with me, should your customers leave you because I did this? Doesn't sound very fair to me. I guess it depends how important I am, doesn't it?
Is Cogent filtering the prefixes they get from Verio? Or is Verio filtering what they send to Cogent? Does it matter?
I think you have a very good point - FT is buying full transit. Cogent is the one without full reachability.
Doesn't mean that FT didn't know this would be a problem when they took the step, though.
Well, FT took the step as you say.. they are the instigator here. But, they are in their right to do so and would have given proper written notice to Cogent so this isnt as much a surprise to them as is being suggested either. Steve
This is part of the game. Party A depeers Party B. Party B has received 30 to 60 days notification. That gives party B enough time to do one of two things. 1) They can ensure they have sufficient transit and/or peering with Party A's customers to ensure that all packets will be delivered. They can make sure that they are seeing all of A's routes through those other networks. This is what's called "being good to your customers" 2) They can take the time to put measures in place to punish party A. Things like route filters to make sure that the only places A gets B's routes are through the (soon to be gone) direct peering. Things like canceling or turning down enough transit so they can claim they CAN'T send the traffic anywhere else. Things like filtering out A's routes from their upstream's BGP feeds. This is called "try to inflict enough pain to effect a reversal" #2 is the stand-off scenario. The next step is, will party B keep up their fortress defense or go home? Will party A back down and turn the peering back up? How long can party B go without their customers canceling? Much of this depends on the relative size of A and B, as well as their customer mixes. If party B is a small porn hoster and party A is a big broadband ISP, then A has nothing to worry about - evidence suggests there won't be many complaints. Customers will be upset, but they won't complain. If party B is hosting something more socially acceptable (if equally pointless) like fantasy football or a popular Everquest server, well, things just got dicey for party A. Party A may just back down. Game theory is fun, folks! With real money on the line, its also very interesting. - Dan On 4/14/05 5:46 PM, "Stephen J. Wilcox" <steve@telecomplete.co.uk> wrote:
On Thu, 14 Apr 2005, Patrick W Gilmore wrote:
On Apr 14, 2005, at 5:16 PM, Richard A Steenbergen wrote:
Surely FT's customers pay for access to Cogents network and vice versa?
In such a case, FT has done its part by paying Sprint for full transit service. It is Cogent who is not accepting the route from their transit, and who intentionally does not carry the global routing table. If I put up a filter on my transit that says I will not accept routes from you unless you peer with me, should your customers leave you because I did this? Doesn't sound very fair to me. I guess it depends how important I am, doesn't it?
Is Cogent filtering the prefixes they get from Verio? Or is Verio filtering what they send to Cogent? Does it matter?
I think you have a very good point - FT is buying full transit. Cogent is the one without full reachability.
Doesn't mean that FT didn't know this would be a problem when they took the step, though.
Well, FT took the step as you say.. they are the instigator here.
But, they are in their right to do so and would have given proper written notice to Cogent so this isnt as much a surprise to them as is being suggested either.
Steve
-- Daniel Golding Network and Telecommunications Strategies Burton Group
dgolding@burtongroup.com (Daniel Golding) writes:
This is part of the game.
more like a war.
Party A depeers Party B. Party B has received 30 to 60 days notification. That gives party B enough time to do one of two things.
1) They can ensure they have sufficient transit and/or peering with Party A's customers to ensure that all packets will be delivered. They can make sure that they are seeing all of A's routes through those other networks. This is what's called "being good to your customers"
2) They can take the time to put measures in place to punish party A. Things like route filters to make sure that the only places A gets B's routes are through the (soon to be gone) direct peering. Things like canceling or turning down enough transit so they can claim they CAN'T send the traffic anywhere else. Things like filtering out A's routes from their upstream's BGP feeds. This is called "try to inflict enough pain to effect a reversal"
for the record, when the old AS174 (PSInet) depeered the old AS6461 (AboveNet), we (old MFN) chose #1. emotionally, i would have preferred #2 -- by a lot! but i could tell that PSInet wasn't going to be alive long enough for this to matter, so there was no long term payoff for the costs of such a war. the customers on both sides of the schism are probably happy that professionalism prevailed... but if PSInet had won and also somehow stayed in existence, then the whole industry would have suffered from the economic power thus conveyed. in other words, sometimes it's better to take pain in a "lump sum" than on the "time payment plan." if that's what cogent's trying to do, they've got my support. if on the other hand cogent is, as accused here today, dumping transit at below cost, then may they rot in hell. (could i say that simpler?)
Game theory is fun, folks! With real money on the line, its also very interesting.
yes indeed. -- Paul Vixie
Paul Vixie wrote:
in other words, sometimes it's better to take pain in a "lump sum" than on the "time payment plan." if that's what cogent's trying to do, they've got my support. if on the other hand cogent is, as accused here today, dumping transit at below cost, then may they rot in hell. (could i say that simpler?)
I'm not sure about the US price war, I just can say that I've seen an offer of AS174 in Switzerland which is 38% of the price of AS1239 we currently pay (same CDR). I'm not sure if ths already justifies hell, but at least purgatory ;-) Regards, Fredy
On Apr 15, 2005, at 2:10 PM, Fredy Kuenzler wrote:
Paul Vixie wrote:
in other words, sometimes it's better to take pain in a "lump sum" than on the "time payment plan." if that's what cogent's trying to do, they've got my support. if on the other hand cogent is, as accused here today, dumping transit at below cost, then may they rot in hell. (could i say that simpler?)
I'm not sure about the US price war, I just can say that I've seen an offer of AS174 in Switzerland which is 38% of the price of AS1239 we currently pay (same CDR). I'm not sure if ths already justifies hell, but at least purgatory ;-)
Does that imply that 1239 has a better line on what "actually costs" are, or else sets those costs for other operators? Tom
On Apr 15, 2005, at 2:10 PM, Fredy Kuenzler wrote:
Paul Vixie wrote:
in other words, sometimes it's better to take pain in a "lump sum" than on the "time payment plan." if that's what cogent's trying to do, they've got my support. if on the other hand cogent is, as accused here today, dumping transit at below cost, then may they rot in hell. (could i say that simpler?)
I'm not sure about the US price war, I just can say that I've seen an offer of AS174 in Switzerland which is 38% of the price of AS1239 we currently pay (same CDR). I'm not sure if ths already justifies hell, but at least purgatory ;-)
Strange, I am REALLY HAPPY when someone offers me a comparable product for less money. If you prefer to pay more, well, I'm happy for you. (And no flames about Cogent not being comparable. I've already posted here that their network runs just fine. Of course, now that they are no longer offering "full transit", we are re-considering how "good" their pricing is.) Back on topic, I am unclear on why "you sell X for less than I do" is justification for rotting in hell. Whether X is below your cost is COMPLETELY immaterial. Whether it is below their cost is irrelevant to me, but it might be important to some countries / laws / whatever. If they are breaking a law, have someone investigate & charge them. If there is no law against it, deal with it. They should be going out of business RSN anyway. -- TTFN, patrick
On Fri, 15 Apr 2005, Patrick W. Gilmore wrote:
On Apr 15, 2005, at 2:10 PM, Fredy Kuenzler wrote:
Paul Vixie wrote:
in other words, sometimes it's better to take pain in a "lump sum" than on the "time payment plan." if that's what cogent's trying to do, they've got my support. if on the other hand cogent is, as accused here today, dumping transit at below cost, then may they rot in hell. (could i say that simpler?)
I'm not sure about the US price war, I just can say that I've seen an offer of AS174 in Switzerland which is 38% of the price of AS1239 we currently pay (same CDR). I'm not sure if ths already justifies hell, but at least purgatory ;-)
Strange, I am REALLY HAPPY when someone offers me a comparable product for less money. If you prefer to pay more, well, I'm happy for you.
(And no flames about Cogent not being comparable. I've already posted here that their network runs just fine. Of course, now that they are no longer offering "full transit", we are re-considering how "good" their pricing is.)
Back on topic, I am unclear on why "you sell X for less than I do" is justification for rotting in hell. Whether X is below your cost is COMPLETELY immaterial. Whether it is below their cost is irrelevant to me, but it might be important to some countries / laws / whatever. If they are breaking a law, have someone investigate & charge them. If there is no law against it, deal with it. They should be going out of business RSN anyway.
The carrier transit market is in a real mess, many of the large players having already gone bankrupt once, when that happened with the so called dot-com crash it affected the global economy. What we see now are todays carriers facing the risk of bankruptcy again, some for the second time round. One possible interpretation of Cogent's price erosion is that networks have been built that will not make a profit until 2 or 3 years and being forced to cut prices heavily in order to compete is pushing that profitability curve such that the companies will run out of funding before they hit profit. The short term benefit as a buyer is reduced costs, but long term this could affect the very market that you're operating in and your own viability and profit margins. For many folks too the falling price they buy transit for just means they are being forced to take that off their product sell prices so they dont actually make any more profit.. in which case there is no advantage to buying below cost services. In general I'd prefer to operate in a healthy marketplace, where all parties are making money, theres little risk of the supplier filing bankrupcty and I am getting reasonable customers service. That can only lead to growth of the industry, healthy businesses and healthy economies. Unforunately none of these things appear to be happening at the moment... Steve
On Sat, 16 Apr 2005, Stephen J. Wilcox wrote:
In general I'd prefer to operate in a healthy marketplace, where all parties are making money, theres little risk of the supplier filing bankrupcty and I am getting reasonable customers service. That can only lead to growth of the industry, healthy businesses and healthy economies. Unforunately none of these things appear to be happening at the moment...
Yes, it's quite interesting what has happened the past few years. I have seen Cisco OC48E/POS-SR-SC cards on the used market sell at $3500 back in 2001/02, go up to $20k in 2004 and they seem to go for around $10-12k now. I also did some other calculations. Let's say the average packet traverses 5 GSR-GSR links and you want to pay the CAPEX over 36 months and let's say on average you fill half those with customer paying traffic. Let's assume you get 50% discount from cisco on these cards and you want to build OC192 links: List price for OC192 cards are $225k, you need 10 of them which with the above equals 225000*10*.5=1125k. At 1125k CAPEX you can carry 5gigabit/s of customer traffic, over 36 months this boils down to 1125000/36/5000=6.25. So just the Cisco linecard CAPEX cost alone for this is $6 / megabit. With the market price seemingly hovering around $15-25 / megabit/month in bulk, this is a large portion of the total cost, considering DWDM systems and people don't come cheap either. So what will people do? Stop selling when their networks are full? Ignore the economics and let other business carry the cost of bulk internet? Go for cheaper platforms? Go bankrupt (if no other business can carry the cost) ? My guess is that going to OC768 isn't going to be very cheap either... -- Mikael Abrahamsson email: swmike@swm.pp.se
Mikael Abrahamsson writes:
So what will people do? Stop selling when their networks are full? Ignore the economics and let other business carry the cost of bulk internet? Go for cheaper platforms? Go bankrupt (if no other business can carry the cost) ?
This problem will be fixed when the excess capacity built in the latter years of the boom is gone. That's not to say that the adjustment won't be painful - I'm sure a few more provider failures are in the offing - but obviously if the marginal price for bandwith doesn't pay for the capital costs of expansion, either eventually bandwidth will be more expensive, or the equipment will be cheaper. Joe
On Sat, 16 Apr 2005 jmalcolm@uraeus.com wrote:
Mikael Abrahamsson writes:
So what will people do? Stop selling when their networks are full? Ignore the economics and let other business carry the cost of bulk internet? Go for cheaper platforms? Go bankrupt (if no other business can carry the cost) ?
This problem will be fixed when the excess capacity built in the latter years of the boom is gone. That's not to say that the adjustment won't be painful - I'm sure a few more provider failures are in the offing - but obviously if the marginal price for bandwith doesn't pay for the capital costs of expansion, either eventually bandwidth will be more expensive, or the equipment will be cheaper.
Hmmmm, router and optical gear capabilities are growing faster than the market. Can you say "permanent state of affairs". moores law > market growth A better (healthier? more sane?) metric is revenue per customer. Mike. +----------------- H U R R I C A N E - E L E C T R I C -----------------+ | Mike Leber Direct Internet Connections Voice 510 580 4100 | | Hurricane Electric Web Hosting Colocation Fax 510 580 4151 | | mleber@he.net http://www.he.net | +-----------------------------------------------------------------------+
On Sun, 17 Apr 2005, Mike Leber wrote:
Hmmmm, router and optical gear capabilities are growing faster than the market. Can you say "permanent state of affairs".
Do you have any facts to back up this statement, as I am of another opinion. We're seeing doubling in traffic growth each year and the optics seem to quadruple every 3-4 years. Of course if equipment cost goes up the vendors will have more resources to make more expensive gear, but I believe the price per bit has to come down rather than the opposite.
moores law > market growth
Well, speaking of that, the performance increase in harddrives and CPUs seem to have degenerated over the past year and the hefty price per bit decrease in harddrive space we saw 1999-2003 seem to have stopped in 2004.
A better (healthier? more sane?) metric is revenue per customer.
Increase in usage is of course driven by the price of bandwidth going down. Let's say for the sake of argument that by 2010 we want to give every household 5 megabit/s on average. How could this be done with technology today seen on the radar? Remember that the households should want to pay for the bandwidth as well, meaning they might be willing to pay $30 per month for the bandwidth part (this is kind of high, but let's go with it). 5 megabit/s at $30 is $6 per meg, with no oversubscription. It also means that for a large city with a million households we need to shuffle around 1M*5M=5Tbit/s in that city alone. We probably need 10Tb national backbone to handle it, how do you do that with 40G or 100G links? That's a lot of parallell DWDM links :/ So the distributed p2p networks of the future will probably have to wait, and instead we'll do regular 1-many broadcasting via these networks and we'll be an interactive cable TV distribution instead. -- Mikael Abrahamsson email: swmike@swm.pp.se
Let's say for the sake of argument that by 2010 we want to give every household 5 megabit/s on average. How could this be done with technology today seen on the radar? Remember that the households should want to pay for the bandwidth as well, meaning they might be willing to pay $30 per month for the bandwidth part (this is kind of high, but let's go with it).
fwiw, 100mb to the home costs about that in japan randy
On Sat, 16 Apr 2005, Randy Bush wrote:
fwiw, 100mb to the home costs about that in japan
Well, I dont really see the average home actually using 100meg all the time in the near future, thus my 5 meg utilization average estimate. Access could be whatever speed of course, access speed not used doesn't cost very much. -- Mikael Abrahamsson email: swmike@swm.pp.se
On Sat, 16 Apr 2005 22:23:53 -1000 Randy Bush <randy@psg.com> wrote:
Let's say for the sake of argument that by 2010 we want to give every household 5 megabit/s on average. How could this be done with technology today seen on the radar? Remember that the households should want to pay for the bandwidth as well, meaning they might be willing to pay $30 per month for the bandwidth part (this is kind of high, but let's go with it).
fwiw, 100mb to the home costs about that in japan
randy
Hi Randy; Do you have any idea what sort of underprovisioning is typical for this sort of service in Japan ? Do they really have anything like a symmetric 100 Mbps all the way back to the backbone ? Regards Marshall Eubanks
Do you have any idea what sort of underprovisioning is typical for this sort of service in Japan ? Do they really have anything like a symmetric 100 Mbps all the way back to the backbone ?
yep randy
On Sun, 17 Apr 2005, Randy Bush wrote:
Do you have any idea what sort of underprovisioning is typical for this sort of service in Japan ? Do they really have anything like a symmetric 100 Mbps all the way back to the backbone ?
yep
Do you have any reference for this? Provisioning 10G distribution links for every 100 customers doesn't seem a viable business model considering the cost of equipment today. Where is the aggregation done and at what traffic level? -- Mikael Abrahamsson email: swmike@swm.pp.se
On Sat, 16 Apr 2005, Stephen J. Wilcox wrote:
For many folks too the falling price they buy transit for just means they are being forced to take that off their product sell prices so they dont actually make any more profit.. in which case there is no advantage to buying below cost services.
To quibble with the economic reasoning here: What it costs you to provide service is dependent on what your costs are (by definition). What you can get your customers to pay you (the market price for your services) is far more dependent on your competitors' costs than on your own. If it costs your competitors less to provide a service than it costs you, they can undercut you. If it costs you less to provide a service than your competitors, you can charge what they charge and make a bigger profit than they do. This means if your competitors are buying something cheaply, it doesn't help you. If you can buy something cheaper, it does help you. But it may only help you keep up, if the same pricing is available to your competitors. The same goes for if your competitors are pretending to have lower costs than they do, although that's generally not a situation that can continue forever. I have no idea whether that's what Cogent is doing; I've seen the claim that bandwidth can't possibly stay as cheap as it is proven wrong too many times at too many steadily decending price points to put much stock in it.
In general I'd prefer to operate in a healthy marketplace, where all parties are making money, theres little risk of the supplier filing bankrupcty and I am getting reasonable customers service. That can only lead to growth of the industry, healthy businesses and healthy economies. Unforunately none of these things appear to be happening at the moment...
Customer service and risk protection may be in better supply than raw capacity. Perhaps there's a market for those services, that might differentiate those who do really well at it from those selling raw capacity cheaply. If not, that's a sign that either those providing cheap capacity are providing service their customers think is acceptable, or the customes don't want good service enough to pay what it would cost. What we're seeing in *some* telecommunications markets -- those once seen as where the money was -- is that there's a huge oversupply of capacity, so it's not worth much. As long as it's not worth much, you aren't likely to see those who spent money on it as if it was worth something making money by selling it. That would seem to be the "fault" of those who built so much of it, or bought it for overly high prices. The good news is that the consumers -- those who are trying to use the stuff rather than just to make money from it -- are benefiting greatly. That's what we're all here for, right? And the bit of good news for those who still want to build this stuff is that there are a lot of places where telecommunications capacity is still in very short supply. Those are generally places without much money, so there might not be all that much money to be made in some of them. But who knows... building lots of identical networks connecting the places with money doesn't appear to have been all that profitable either. -Steve
Cogent is now reachable from OT and vice versa, apparently Cogent dropped the filters, i see everything passing verio now. Not sure since when this works again. Regards, Jonas
For many folks too the falling price they buy transit for just meansthey are being forced to take that off their product sell prices so they dontactually make any more profit.. in which case there is no advantage to buying below cost services.
In recent years, the unregulated telecoms industry has struggled with the steep slide towards commoditization. This is a problem because the industry's definition of telecom services is so narrow that there are few opportunities to add value and outrun the commoditization wave. Moving packets from point A to point B is rapidly becoming as glamourous and as profitable as moving water from point A to point B or moving gas, or moving electricity... The only way out is for the telecoms industry to be dismantled by 3rd parties who will buy and own networks in order to leverage those networks for their own value-add services. Everyone else will just have to get used to repeated cost-cutting exercises. --Michael Dillon
Patrick W. Gilmore wrote:
I'm not sure about the US price war, I just can say that I've seen an offer of AS174 in Switzerland which is 38% of the price of AS1239 we currently pay (same CDR). I'm not sure if ths already justifies hell, but at least purgatory ;-)
Strange, I am REALLY HAPPY when someone offers me a comparable product for less money. If you prefer to pay more, well, I'm happy for you.
Product is not comparable anymore, you'll miss ^174_5511_ ... just kidding.
Back on topic, I am unclear on why "you sell X for less than I do" is justification for rotting in hell. Whether X is below your cost is COMPLETELY immaterial. Whether it is below their cost is irrelevant to me, but it might be important to some countries / laws / whatever. If they are breaking a law, have someone investigate & charge them.
Not in the so called free markets like US or Switzerland, right? The real horror comes when connecting comanies and enterprises away from the main street. When local tail can only be provided by the incumbent. How to make a good deal for the customer (IP feed + local tail) when the IP Transit price is probably less than 2% of the actual costs? I'd be quite pleased with AS174's price dumping policy, at least it helps to beat down the price of other suppliers and makes us more competitive. F.
On Thu, Apr 14, 2005 at 05:29:46PM -0400, Patrick W Gilmore wrote:
Is Cogent filtering the prefixes they get from Verio? Or is Verio filtering what they send to Cogent? Does it matter?
Or OT tagging their announcements to Sprint in a way that prevents them being announced to Cogent in order to force Cogent into buying transit.
I think you have a very good point - FT is buying full transit. Cogent is the one without full reachability.
Indeed. That's the inherent danger of "playing tier 1". I myself find being tier 2 much more comfortable. :-) I'm not sure wether the effort to become (and maintain!) tier 1 status (and I'm talking of the technical term == transit free) is economically a good idea (anymore).
Doesn't mean that FT didn't know this would be a problem when they took the step, though.
I wonder how much non-BGP-transit traffic is on Cogent (static customers, colocation). My guess is that it's quite low and one might say "well, BGP routes around the problem". People do have multiple upstreams (often seen in the setup "primary Cogent because of price, and 1-2 others secondary for quality and resilience"). Best regards, Daniel -- CLUE-RIPE -- Jabber: dr@cluenet.de -- dr@IRCnet -- PGP: 0xA85C8AA0
On Fri Apr 15, 2005, Daniel Roesen wrote:
Or OT tagging their announcements to Sprint in a way that prevents them being announced to Cogent in order to force Cogent into buying transit.
For people interested hereafter our route-server: telnet://route-server.opentransit.net German
On Thu, Apr 14, 2005 at 06:52:49PM -0400, German Martinez wrote:
Or OT tagging their announcements to Sprint in a way that prevents them being announced to Cogent in order to force Cogent into buying transit.
For people interested hereafter our route-server:
telnet://route-server.opentransit.net
That doesn't show what your outgoing route-map/policy-chain looks like towards your Sprint hand-offs. :-) Just to make sure: I'm NOT implying you do that tagging, I just say that it's a technical possibility which wasn't mentioned. Best regards, Daniel -- CLUE-RIPE -- Jabber: dr@cluenet.de -- dr@IRCnet -- PGP: 0xA85C8AA0
On Fri, Apr 15, 2005 at 12:36:22AM +0200, Daniel Roesen wrote:
On Thu, Apr 14, 2005 at 05:29:46PM -0400, Patrick W Gilmore wrote:
Is Cogent filtering the prefixes they get from Verio? Or is Verio filtering what they send to Cogent? Does it matter?
Or OT tagging their announcements to Sprint in a way that prevents them being announced to Cogent in order to force Cogent into buying transit.
Scratch that. I was under the (mis-)assumption that Sprint peers with Cogent. Best regards, Daniel -- CLUE-RIPE -- Jabber: dr@cluenet.de -- dr@IRCnet -- PGP: 0xA85C8AA0
Is Cogent filtering the prefixes they get from Verio? Or is Verio filtering what they send to Cogent? Does it matter?
I think you have a very good point - FT is buying full transit. Cogent is the one without full reachability.
Doesn't mean that FT didn't know this would be a problem when they took the step, though.
Cogent has offered to exchange traffic with OT in a manner that is cost-effective for both sides. OT said, "no, do it in a way that is more expensive for both of us". Which one is making a reasonable effort to exchange traffic with anyone else willing to make a similar effort? To me, that is the standard by which connectivity should be judged. DS
If I may, you sound like someone whom FT has depeered in the past? :)
Personally no. ;) simply playing devils advocate - who really knows what business model people are following? who really knows why this has happened? But in my view this type of action where it impacts customers doesn't help any of us. If you can't compete at the market price then you shouldn't be in the business, aggressive pricing in the market is just another thing you need to deal with. Regards, Neil.
from FT: FT terminated its direct interconnection with Cogent as they did not comply with 2 of the criterias of the FT Peering Policy. This Policy is official and published. However, route exchanges between Cogent and FT customer remained possible through IP transit provider networks such as Sprint for FT. As a matter of retaliation, Cogent "blackholed" all FT IP addresses, cutting the routes between their customers and FT's ones. This unitaleral measure is a breach of the rules agreed upon among the IP community. FT cannot be held responsible vis a vis its customers for the negative consequences of an action taken under the sole responsibility of Cogent. Regards, Jonas
This is a matter of human nature, I suppose. Everyone is terribly pleasant when they hear what they want. The true test is what happens when folk hear the "wrong" answer. I've depeered and I've been depeered. I've seen folks on the receiving end of bad peering news handle it with consummate professionalism. I've also seen folks act like spoiled children, forgetting the fundamental rule of peering: Peering is a business relationship. Peering is about meeting the mutual business needs of two networks. Emotionalism , "hurt feelings", and actions that violate the bounds of trust and the normal bounds of professionalism have no place in internetwork peering. Depeering is always a gamble and, as such, is to be generally avoided as unnecessary risk. Given that, folks need to resist their urge to put black hats on networks who decide that certain peering relationships have outlived their usefulness. The true picture is always more complex than the spaghetti western. If enough folks are actually interested, I'd be happy to do a talk at an upcoming NANOG on depeering (methods, etiquette, likely outcomes, necessary pre-action analysis). This might be good for a future peering track. - Dan On 4/14/05 1:38 PM, "Steve Gibbard" <scg@gibbard.org> wrote:
On Thu, Apr 14, 2005 at 11:28:00AM -0400, dgolding@gmu.edu wrote:
in depeering. However, dealing with Cogent on peering matters is incredibly unpleasant. I can understand networks and peering coordinators feeling that it just isn't worth it.
Just for the record, I've dealt with Cogent's peering people on behalf of a few networks over the last two years, and in my experience they've been extremely pleasant to work with.
-Steve
participants (23)
-
Alexander Koch -
Daniel Golding -
Daniel Roesen -
Daniel Senie -
David Schwartz -
Deepak Jain -
Fredy Kuenzler -
German Martinez -
jmalcolm＠uraeus.com -
Jonas Frey -
Marshall Eubanks -
Michael.Dillon＠radianz.com -
Mikael Abrahamsson -
Mike Leber -
Neil J. McRae -
Patrick W Gilmore -
Patrick W. Gilmore
-
Paul Vixie -
Randy Bush -
Richard A Steenbergen -
Stephen J. Wilcox -
Steve Gibbard -
Tom Vest