ratios

Mailman List Mirror (Read Only)
1

Peter Jansen wrote:

Scott:
Have a look at our peering policy at www.cw.com/peering. It will
provide you with some information on peering with large networks.

This should read: "Have a look at our peering policy at www.cw.com/peering
if you want to see a prime example of how *not* to develop your peering
policy."

Peter, I can't believe you have the testicular fortutude to come on this
list with this garbage. Do you think we have forgotten the PSI/C&W peering
fiasco? Dude, you had *paying* customers who depended on having routes
into AS174, and you turned your back on them, knowing damn good and well
that PSI was in no position to purchase transit. In fact, I'm surprised to
see that you're still a C&W employee after all that mess.

Plus, wtf is this clause about announcing 5000 routes? What a crock of
s**t! This really encourages aggregation, doesn't it?

You'd think, after all of the Exodus customers jumping ship after C&W
bought them, that you'd start rethinking your business practices. People
want, and are willing to pay for, a well connected network, and AS3561
isn't, at least not to the outside world.

Also, I would like to point out that if you're mostly content, I doubt
you'll ever be able to meet C&W's peering criteria because they have
practically no eyeballs. Their wholesale dial division is pretty much
extinct, and they seem to be leaning more toward hosting than selling T1s
to mom and pop dialup ISPS.

It's a great big catch-22 any way you look at it, and I hate you, Peter
Jansen, for it. There's a special level of hell for people like you when
you die.

Dean

2

</lurk>

  Well, Never Forget -the- man who defined peering to the FCC,
among other american institutions, in modern times .....

  is no less than "Bernie" of WorldCom.

  And, as "Bernie" leaves us, facing allegations of all sorts of things..

  Let us never forget -he- was selected to be the "Ethical"
  guiding light for the internet, circumscribing the boundaries
  of our .... policies.

  And CW's policy is merely following in his footsteps,
  following "precedent", as it were. (IMHO)

  Of course, I, for one,
   was never one to follow the Pied Piper....

  As PSI, Global Center, Enron, The Leader of Earthlink....
  and others, were to learn the wisdom thereof.

  An Internet -without- peering, hrmmm......
    
  FWIW, I noticed a trend of people withdrawing peering
  shortly before chapter 13'ing.

   * shrug *

  BTW, anyone checked the value of WorldCom stock recently ?

  "No man is an island"

<lurk>
   
Dean S Moran wrote:

3

It's a great big catch-22 any way you look at it, and I hate you, Peter
Jansen, for it. There's a special level of hell for people like you when
you die.

a little harsh :wink:

4

Plus, wtf is this clause about announcing 5000 routes? What a crock of
s**t! This really encourages aggregation, doesn't it?

And even AS6461 barely squeaks by with 5571 routes the last time I checked
a couple weeks ago. I don't think this policy is for real - if they
actually enforce it then it will completely change the tier-1 landscape.
Here's few more stats I just checked:
Verio AS2914 - 1430 prefixes
L3 AS3356 4168 prefixes
Genuity AS1 - 7406 prefixes

-Ralph

5

I have some trouble seeing why folks are so interested in meeting or
debating peering requirements set out by carriers that have made it quite
clear that they are not taking new peers. Most of the published requirements
from these carriers serve two functions - to prevent new peers, and to
depeer those who are felt to be not worthy. And even the latter is tenuous -
most bilateral peering agreements allow for cancellation at will for
absolutely no cause.

Peering is a business relationship. Refusing to peer does not make one bad,
nor does it damn the peering coordinator to eternal damnation. It also does
not reflect on those who work for the carrier in other roles, especially
those brave enough to post to NANOG on peering matters. Some folks take
exception to having ANY sort of peering requirements, like the person who
told me that they thought a carrier that required bicoastal peering and an
OC-12 network has peering requirements "worse than UUNET". Peering
requirements, especially rational ones like multiple location peering, are
not in any way bad.

If you don't approve of a carrier's peering policy, you have a couple
options...

You can publicly denounce them on a forum like this, which has doubtful
effect.

You can turn away their sales folks, the next time they try to sell you
transit. However, if you say "I won't buy transit from you, because you
won't peer from me", don't expect any sort of reaction other than "goodbye",
because there is no lost revenue potential - you would never have purchased
transit in any case. However, if you say "because you won't peer with other
large networks, it decreases the quality of your network, so I won't buy
your transit". They may be more effective. However, that needs to happen
much more than the sales people hear "I won't buy transit from you because
I'm a peer".

You can take it out on individuals who you feel are responsible, by refusing
to do business with them or hire them in the future. This is very tricky, as
all employees of a carrier are not in any way responsible for a carrier's
peering policy. Of course, if you get some weasel who comes in for a job
interview, with "senior peering engineer" on their resume, and brags about
his role in depeering, say, PSI, then I suppose such persons deserve what
they get. However, it's rare that this comes up. Additionally, punishing
folks for enforcing rational peering requirements is counterproductive.

I guess the best thing you can do is not take peering matters personally,
and to remember that peering decisions are business decisions, and they by
personalizing them, it creates unnecessary animosity.

- Daniel Golding

6

I actually think this is put very well. I know that in my case I'd
prefer to buy transit from a company who has an open peering policy.
For example, I'd certainly consider buying transit from mfn before uunet
for example. I realize there are many other factors including
relyability, cost, company stability etc. but one consideration ior me
is their willingness to peer and grow their networks. I wouuld think
especially on this list our arguments should stick to being as strictly
technical as possible and not venture in to the personal. Easier said
than done I realize. However, strong arguments for using networks with
open peering policies are more meaningful than ridiculing large carriers
who don't wish to peer.

The only thing I can say is I wish they would just publically
acknowledge that fact. If uunet and cw don't wish to peer they should
just not have a peering policy.

7

Oh come now Dan, that is too much of a logical decision for this mailing list :slight_smile:

-Steve

8

I have some trouble seeing why folks are so interested in meeting or
debating peering requirements set out by carriers that have made it quite
clear that they are not taking new peers. Most of the published requirements
from these carriers serve two functions - to prevent new peers, and to
depeer those who are felt to be not worthy. And even the latter is tenuous -
most bilateral peering agreements allow for cancellation at will for
absolutely no cause.

Most that I've signed require a bit more notice than, say, a couple days. Ours requires 60 days, unless it's for some cause ( which still provides a period for the other network to "cure" ).

You can publicly denounce them on a forum like this, which has doubtful
effect.

It informs other networks of the actions taken by said carrier. Other
networks may in turn change *their* business decisions based on that
information.

Additionally, punishing folks for enforcing rational peering
requirements is counterproductive.

Rational is a pretty subjective concept. :slight_smile: Overly restrictive
covenants wrt housing have been struck down in the past. One could
make an analogy to overly restrive peering agreements if one wanted
to go down that particular rat hole...

I guess the best thing you can do is not take peering matters personally,
and to remember that peering decisions are business decisions, and they by
personalizing them, it creates unnecessary animosity.

Aye, and just like any business decision they affect other business
decisions.

Regardless, it smarts a bit when a peer with no warning terminates
peering and then offers to sell you transit in the same breath. It
smacks a little too much of trying to strong-arm the peer into having
to pay for transit by not giving them sufficient time to develop alternate
arrangements.

-Chris

9

I imagine "public relations" depts of CW and UUNET will not allow them to
just admit they would not peer, and this is why they "have" peering policies.

In reality however, UUNET will peer with anyone who pays them money for
the peering traffic and they do provide good discounts on this. Add to that
some ISPs have negotiated such that they get transit as part of their
uunet peering arrangements. I don't have know about CW currently, have
not dealt with them for long time, but years ago CWUSA (before buying MCI)
did peer willingly and MCI never did really. One interesting note is that
UUNET does peer with CW itself in number of new locations, for example big
peering point for both of them is Equinix in San Jose. Since I can't
imagine those companies going there just to peer with each other, they
must have number of other peers their as well... Equanix people - do you
want to comment on this?

As for buying from companies that are more open to peering, I think this
should come into play only if everything else is equal and more important
maybe their network and their willingness to work with customer on BGP and
setup custom filters, communities, etc.

10

I actually think this is put very well. I know that in my case I'd
prefer to buy transit from a company who has an open peering policy.
For example, I'd certainly consider buying transit from mfn before uunet
for example. I realize there are many other factors including
relyability, cost, company stability etc. but one consideration ior me
is their willingness to peer and grow their networks.

The status of those peering links is most important than if they exist or
not. I'm not aware of any major UUNet peers being congested, as compared
to say a company in an extremely difficult financial situation who just
laid off almost all of their engineers and doesn't have any time or
resources for upgrading congested peers.

Yes UU only privately peers with like 20 people and it's quite possible
that your traffic will just go to another large backbone and be congested
there, but thats the chance you take in this world which believes in
claiming "99.999% uptime with self healing sonet rings!" instead of
showing you their actual capacity and utilization.

If you choose to peer with only 20 people, but do it with massive amounts
of capacity, that is one thing. If you keep old congested peers around,
and you refuse to upgrade them or stonewall for years promising upgrades
but never follow through because you don't like someones ratio, that is
another. At the end of the day I can still have respect for UU because
they get the packets through, even if I don't necessarily agree with where
they send them to do it.

The only thing I can say is I wish they would just publically
acknowledge that fact. If uunet and cw don't wish to peer they should
just not have a peering policy.

I think Level 3's is still the most difficult on paper, despite the fact
that most of their peers will never meet these requirements
(http://www.level3.com/1511.html):

Dual OC48 into every city
Presence in 15 major cities
1000Mb/s minimum traffic exchanged
Must peer at OC12 or higher
Must peer in 8 locations

11

No, thats why they went there (well that and to sell transit). :slight_smile:

It's a lot cheaper and easier to get a crossconnect done within 48 hours
then it is to get a metro OC12. Multiply that by the number of people they
do peer with, and it adds up to a lot.