Cable & Wireless "de-peering"?!?

Mailman List Mirror (Read Only)
1

The number of customers affected is unimportant, what matters is the
amount of traffic affected. I don't know about anyone else, but I think if
I had just been "de-peered" the provider in question would be the LAST on
my list to purchase transit from. In all likelihood the traffic is just
going to go to another CW peer and to an FNSI transit. But to determine
the true loser, you must know if this peer served a useful technical
function. If this was a low-quality peer (congested, through a lossy atm
nap, etc) or relieved no congestion elsewhere, the loser is FNSI. If on
the other hand this peer was providing a better path, the traffic will be
affected. Since billing is based on traffic, the loser is whoever can no
longer bill their customers for something they got for free.

Also, not that I care much about either FNSI or Clueless & Witless
peering, but the argument that noone would be affected AND traffic would
be reduced makes no sense. If there is a significant reduction in
congestion then there must have been a significant amount of traffic
flowing through the peer. As far as I'm concerned, the biggest argument
for peering with FNSI is Pimp War (http://www.pimpwar.com). :stuck_out_tongue:

2

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

And now, to "drop the bomb"...

The other issue is that this isn't just C&W vs FNSI. The word on the
street is that C&W is engaged in a "Kamikazee Peering" strategy in
order to boost transit revenues. Rumor has it that other recipients
of the C&W nasty-gram include Level 3, Telia, NetRail, XO, and Verio.
There are certain to be others as well. I can't swear on a stack of
bibles that these folks have gotten the letter, but that's the story,
at least. If true, then we are talking about a reasonable chunk of
the commodity Internet.

It's unusual to "name names" in the peering game, but this is an
intolerable and almost unpresidented situation, as some or all of the
networks listed above are transit-free. And, contrary to Randy's
comments, this is certainly enough of the Internet to cause
disruption.

We must ask:

1) Is this the end of peering as we know it? Is it settlement time
all around?

2) Is C&W simply flaking out? Will they end up screwing themselves?

3) Will we stand up to the bully on the playground?

4) Is this a crisis effecting global reachability?

#1 is doubtful - the only current condition that could really cause
this is the collapse of multiple transit-free carriers due to current
market conditions. It would take more than a PSINet implosion to
cause this, I'm guessing.

#2 - C&W's peering decision makers are certainly taking a risk, here.
It is devoutly hoped that their poor judgement comes back to haunt
them. On the other hand, perhaps they'll experience a change of
heart, and be filled with the spirit of Internet Brotherhood.
Needless to say, I'm not holding my breath.

#3 - Well, I doubt it. At least one ISP has already crumbled and
agreed to a settlement with C&W.

#4 - Yep. C&W's actions and the resulting disruptions in service that
we'll see because of them, over the next 30 days, mark a new chapter
in peering disputes. Should be interesting to see how it all pans
out.

Now, back to the regularly scheduled off-topic threads concerning
telcom legislation and blackholing China...

- - Daniel L. Golding (Speaking only for me)

3

I feel I have to comment on the second point of your e-mail. It makes clear points as the how each provider is going to take full responsibility for their peering decisions.

I have seen a trend where Tier 1's (nation-wide large carriers - don't want to debate the term too much) are willing to offer transit contracts to regional or tier2 nationals at cost, or only at a very small margin. This move by C&W to begin charging for those providers seems to be a very large departure from this method.

As we all know, the value of the big player, is how many people they are connected to. It seems that providers are willing to take a a risk by either peering or selling very cheap transit in order to appear as a well connected ISP to the rest of the Internet.

My concern now, after reading this, is how do I know value the ISPs that I buy transit from. What is the guarantees that I have that the bandwidth I expect from a transit carrier is going to be maintained.

I have been a long time customer of C&W well into the MCI days. While I have had my difficulties, I feel that they have been a solid carrier. However, it has been a struggle lately to keep more and more traffic on their backbone. This whole thread seems to be the cause.

In summary, my question is this:

  Does a peering policy like this make sense?

My answer to this is: It may, but it is forsaking two different types of customers. The first, the tier 1/2 customers that is now enforcing a tariff against. The second, regional carriers that can no longer afford to purchase second rate transit.

So, how can I justify paying C&W for this bandwidth that day by day may be losing value.

I'd be interested to hear how many people that feel that C&W's decisions affects their direction for connectivity.

Tony

4

It would be interesting to note if the C&W policy change is reflected
in a quantifiable deterioration of 3561's placement on the CAIDA skitter
chart.

I would like to think that Draconian peering policies leads to economic
pain. It seems to be worthy of note that AGIS and PSI tried to use the
"pay me don't peer" card before they deteriorated. It seems in some
cases more an act of financial desperation rather than acting in the
best long term interest of the network or the customers.

The more open networks move quickly to center stage while the large
networks build walls in moves of quite desperation. 3561 was once the
"North Star" of the Internet constellation. IMHO said stringent
peering policies moves 3561 in a direction that is not in the best
interest of the C&W customers or the C&W stockholders.

   "... the true value of the Internet is its connectedness ..."

5

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

And now, to "drop the bomb"...

Sorry, didn't see this because you're posting from an open relay
listed in RSS, so it went into a spam folder.

You might want to talk to your mail provider or something.
(see ERS - Home Page  | Trend Micro Service Central)

The other issue is that this isn't just C&W vs FNSI. The word on the
street is that C&W is engaged in a "Kamikazee Peering" strategy in
order to boost transit revenues. Rumor has it that other recipients
of the C&W nasty-gram include Level 3, Telia, NetRail, XO, and Verio.
There are certain to be others as well. I can't swear on a stack of
bibles that these folks have gotten the letter, but that's the story,
at least. If true, then we are talking about a reasonable chunk of
the commodity Internet.

Yes, this is a different game.

#3 - Well, I doubt it. At least one ISP has already crumbled and
agreed to a settlement with C&W.

Shame.

#4 - Yep. C&W's actions and the resulting disruptions in service that
we'll see because of them, over the next 30 days, mark a new chapter
in peering disputes. Should be interesting to see how it all pans
out.

It would be nice if this action did end up hurting C&W more than
any other individual provider... will it? Depends on who crumbles

6

Didn't UUNet try this back in 96? A quick search of Boardwatch failed to find the article, but ISTR that John Sidgemore eventually slunk back to the playground and agreed to play nice. If UUNet couldn't pull it off back then, I doubt that CW can now. Things have changed a lot in 5 years, but I would suspect that "Stealing the Internet" would now be harder rather than easier.

7

albert@waller.net (Albert Meyer) writes:

Didn't UUNet try this back in 96? A quick search of Boardwatch failed to
find the article, but ISTR that John Sidgemore eventually slunk back to the
playground and agreed to play nice. If UUNet couldn't pull it off back
then, I doubt that CW can now. ...

I am completely fascinated by your assessment (that UUNet didn't pull it off).

8

Altert:
  Pardon my ignorance on this issue, but I read C&W's receint peering
agreement and it seems they are simply trying to define what is a true
"peer". My question is, where is the traditional line that defines who
should be a "peer" and who should be a customer and shouldn't that be open
to re-evaluation as the network evolves. Is it that C&W has "rigged" their
peering agreement with specifics that would de-peer legitimate peers or
are they just trying to protect themselves from an evolutinary change.

Chuck

9

It is rare, but I agree with Paul here :slight_smile:

Unet is, for example, one of the few (if not the only) ISP in The
Netherlands that charges for *peering* (no, not transit, just peering).

More and more clued people I know are avoiding UUnet because they
don't peer with the small but quickly growing ISPs. Most UUnet
customers are getting worse and worse connectivity as other ISPs stop
peering with UUnet, because UUnet is becoming less and less important.
A nice downward spiral.

Greetz, Peter.

10

I contend that if quantifiable evidence exists that setting the peering
bar high in the name of selling transit and/or restricting new players
leads to a downward spiral in network quality ... then evidence exists
that can be taken to stockholders and BOD of those companies.

Companies, unless they are run by less than competent management, will
not knowingly commit economic suicide. Stockholders tend to have a very
low tolerance for management stupidity. Stockholder tolerance now is
a lot lower then five years ago.

11

[snip]

Unet is, for example, one of the few (if not the only) ISP in The
Netherlands that charges for *peering* (no, not transit, just peering).

I stand corrected on this bit. UUnet just has quite strong peering
requirements. Anybody wanting to exchange traffic with UUnet while
being unable to comply must become a customer, therefore.

More and more clued people I know are avoiding UUnet because they
don't peer with the small but quickly growing ISPs. Most UUnet
customers are getting worse and worse connectivity as other ISPs stop
peering with UUnet, because UUnet is becoming less and less important.
A nice downward spiral.

I do still think UUnet is in a downward spiral, just like C&W. Strong
peering policies are not good for the Internet.

Greetz, Peter.

12

Refusing to peer is not always a bad option if you have a big enough share
of the market. In Australia the top 3 providers (Telstra, Optus,
connect.com) own about 80% of the market, host most of the domestic
content and supply circuits to most other ISPs. It is "pretty hard" to
setup peering with any of them.

They have very little incentive to peer with smaller ISPs. The small
provider will have to connect with one of them anyway and they will
obviously make more money changing this provider for the circuit (and the
traffic) than just peering for free.

On the other hand in New Zealand we have all except on of the large
providers well connected by a couple of peering exchanges. This provider
is under a lot of pressure to peer with everyone else and at least one
other provider has recently shutdown a circuit that was being used for
peering (The large provider is trying to force others to pay for circuits
directly to it, plus some other fees).

13

I do still think UUnet is in a downward spiral, just like C&W. Strong
peering policies are not good for the Internet.

Greetz, Peter.

  I'm not so sure that's true. Personally, I think the Internet is better
served by having a smaller number of larger and better maintained meeting
places than by just having a large number of peering points were everyone
connects to everybody else.

  Here are just a few reasons why, for example, it's better if you use
transit to FooNet to reach BarNet rather than BarNet peering with you
directly (assuming you are not too big yourself):

  1) FooNet and BarNet are more likely to keep their peering points scaled to
handle the load than you are. They are more likely to monitor performance
and shut down failures.

  2) FooNet and BarNet will meet at more places than you will meet BarNet,
allowing traffic to get off the source network faster and providing better
fault tolerance.

  3) Fewer BGP sessions means faster convergence and less instability.

  4) You may be more likely to meet BarNet at public peering points while
FooNet is more likely to meet BarNet at private peering points. Your traffic
to BarNet will get the benefit of the higher amounts of effort FooNet and
BarNet will put into keeping their meeting points efficient.

  DS

14

People made similar claims regarding why AT&T shouldn't have been broken
up... also called the "a few big companies can take care of consumers
better" argument. Watch CSPAN, you'll hear this argument periodically
from the megacorp of the week making it's case before congress.

Historically, a few big companies controlling any particular industry
results in slower innovation until some chaotic event occurs that allows
new entrants that don't have a vested interest in the status quo.

Remember how MCI got its start. They weren't allways that big. Did they
have a right to exist when they where small? AT&T didn't think so and
didn't want them connecting to the phone network. MCI went to court and
the rest is history.

> I do still think UUnet is in a downward spiral, just like C&W. Strong
> peering policies are not good for the Internet.

> Greetz, Peter.

  I'm not so sure that's true. Personally, I think the Internet is better
served by having a smaller number of larger and better maintained meeting
places than by just having a large number of peering points were everyone
connects to everybody else.

  Here are just a few reasons why, for example, it's better if you use
transit to FooNet to reach BarNet rather than BarNet peering with you
directly (assuming you are not too big yourself):

  1) FooNet and BarNet are more likely to keep their peering points scaled to
handle the load than you are. They are more likely to monitor performance
and shut down failures.

  2) FooNet and BarNet will meet at more places than you will meet BarNet,
allowing traffic to get off the source network faster and providing better
fault tolerance.

  3) Fewer BGP sessions means faster convergence and less instability.

  4) You may be more likely to meet BarNet at public peering points while
FooNet is more likely to meet BarNet at private peering points. Your traffic
to BarNet will get the benefit of the higher amounts of effort FooNet and
BarNet will put into keeping their meeting points efficient.

  DS

+------------------- H U R R I C A N E - E L E C T R I C -------------------+

15

peter@dataloss.nl (Peter van Dijk) writes:

I do still think UUnet is in a downward spiral, just like C&W. Strong
peering policies are not good for the Internet.

Well, before I returned to PAIX, I was CTO of MFN. MFN had bought Abovenet,
and I learnt to respect strong peering policies. Abovenet's peering policy
was, and as far as I know still is, very "strong". It's strongly _open_,
which means it's different from the purported C&W policy being discussed here
this week. But while different it is still "strong".

I think what you were looking to say above is that "*closed* peering policies
are not good for the Internet." Probably noone will *publically* disagree
with such sentiments. Not on nanog, anyway.

16

I do still think UUnet is in a downward spiral, just like C&W. Strong
peering policies are not good for the Internet.

Is this just an opinion, or do you have evidence?

A genuine impartial economic analysis of this issue would be
very interesting, and probably even more interesting to the
regulators. I would hope any such analysis will include cost
of maintenance and stability of networks with n interconnects
as n increases.

Most regulation theory looks to minimize customer price
over the long term. This does not necessarilly mean maximizing
competion (though it normally but not always means no monopolies).

European regulation in this area (yes it exists) focusses only on those
with Significant Market Power (variously defined but normally
25% or above of the 'Relevant Market'), and suggests that their
interconnect policies should be non-discriminatory and cost-oriented.
In my book that means that A has 30,000 routes of 100,000, and B has
2 routes, then B is paying A. (before saying that 'but B would
charge A too' remember that A probably has the opportunity to
exchange data via B's transit, at, no doubt, a lower cost per bit).
In the mean time their view is, in general, that if what you
say is true (downward spiral etc.) because their peering
policy is suboptimal, then this just allows some competitor
with a more optimal peering policy to gain competitive advantage.

IMHO there is a high correlation between people claiming
strong peering policies are 'bad for the internet' and
those recently refused peering. Thus data samples are
somewhat skewed.