Peering - Benefits?

Mailman List Mirror (Read Only)
1

Hi there...

I'm in a meeting next week to discuss settlement-free peering etc.....
always an interesting time. A push is on (by myself) to get into other
physical locations and participate on the peering exchanges.

Besides costs, what other factors are benefits to peering?

I can think of some but looking to develop a concrete list of appealing
reasons etc. such as:

-control over routing between networks
-security aspect (being able to filter/verify routes to some degree)
-latency/performance

Just looking for other positive ideas etc...:wink:

Cheers!

Paul

2

I'm surprised you didn't include "chance to pick up a redundant connection".

3

Thanks! That's a really good one and surprised myself I missed it..:wink:

4

It would only be a redundant connection if the AS your peering with is a
transit AS. The AS that I work with is a stub AS and can not function as
a fully redundant link.

Just something to watch out for.

Paul Stewart wrote:

5

Thanks - I believe the wording meant was "alternative path" versus
connection... in other words if an AS has issues with one or more
upstream providers for whatever reason, you have good chances the
peering connection will remain in better shape (not always granted, but
good odds)....

Paul

6

But if that AS is a stub, you still can't use their up stream providers
to get data out to the rest of the world. It still wouldn't even
function as an "alternative path" it would only function for the subnets
which that AS owns.

Paul Stewart wrote:

7

allows geeks to go on junkets almost as cool as droids get

8

...specifically, in non-carrier-owned colos you have a better chance of
factoring out loop costs for pricing decisions.

A couple to add:
- failure scoping: issues on a remote network can be better isolated
  from the rest of your traffic (or completely if it is the peer).
- product variation: if you sell connectivity, a different/diverse/rich
  set of paths to offfer your downstreams is a win.

9

Joe Provo wrote:

A couple to add:
- failure scoping: issues on a remote network can be better isolated from the rest of your traffic (or completely if it is the peer).
  

Related to this is ability to contact the right people more quickly. If you've got a problem with/on someone's network then typically you can call their NOC directly. Compared with having to bounce through your transit providers helpdesk, who then escalate to their NOC, to the other NOC etc. This right is usually enshrined in most people's peering policy requirements.

It's a powerful thing and not to be underestimated.

MMC

10

internet exchanges are not per-se "redundant"
they basically are a switch which actually, because of the many connected
parties, most of which do not have enough PAID transit to cover any
outages on it, causes more problems than they are good for.
(the amsix with their many outages and connected parties that rely
primarliy on it's functionality is a prime example here)

internet exchanges usually are some sort of hobby computer club, you
cannot rely on them to actually -work-, but when they do work that's
"nice" (always make sure you have enough paid capacity to cover for it
when they do not work however!)

peering on only one of them therefore does not make your network more
reliable in any way (it becomes a different story when you connect to like
10 or so worldwide).

as for "peering" agreements, just implement an open peering policy
(doesn't nessesarily have to take place over an ix, also applies to pieces
of ethernet running from your network to others).

those basically are contracts that force anyone who has also signed one to
peer with your network, wether they like you or not (saves the trouble
when you are a content provider and others do not want to peer with you
because they provide content too and you are a competing party etc).

11

Thanks - no I understand that...

We have multiple transit providers today and are already present on a
couple of smaller peering exchanges with an open peering policy... our
experience with them has been very positive.

The redundancy perspective is that you now have more paths to the same
AS - and an assumption that the peering route will always be best (I
know that's not always true). We of course have enough transit in case
of a peering outage - would never "put all our eggs into one basket"
that it sounds like some others are doing.... also, we are looking at a
number of them in various parts of the world currently which adds
another level of redundancy per say....

Take care,

Paul

12

HRH Sven Olaf Prinz von CyberBunker-Kamphuis MP wrote:

as for "peering" agreements, just implement an open peering policy
(doesn't nessesarily have to take place over an ix, also applies to pieces
of ethernet running from your network to others).

those basically are contracts that force anyone who has also signed one to
peer with your network, wether they like you or not (saves the trouble
when you are a content provider and others do not want to peer with you
because they provide content too and you are a competing party etc).

It is not practice in this community for 'open peering policy' to mean 'must
peer with anyone'. You might still refuse to peer on the basis that the other
party is unreliable or run by idiots, and this is perfectly acceptable even
with an advertised open peering policy.

Nor does such a statement create any form of contract or obligation under any
law I am aware of, as such an indicative offer does not fulfill the
requirements to form a binding contract.

Any device which has REQUIRED e.g. participants in an IX to peer with others
has proved very unpopular in the industry.

13

internet exchanges are not per-se "redundant"

Those networks who *choose* connect to peers via a single fabric, in a single location, will suffer a similar fate to those networks who single home to one transit provider.

(the amsix with their many outages and connected parties that rely
primarliy on it's functionality is a prime example here)

I run interconnection for three networks connected to the ams-ix and can't understand why you think that the ams-ix fabric has "many outages" - I have found it, to borrow a phrase from another stable European IXP, rock solid.

internet exchanges usually are some sort of hobby computer club, you
cannot rely on them to actually -work-,

You shouldn't bet the farm on a single one of anything. My alarm clock failed once, this doesn't mean that all alarm clocks are hobbyist timekeeping devices.

Most internet exchanges are professional organisations run by a team of experts who care about the operational stability of the platform. Most in Europe are association based organisations, who's leaders are held accountable for the operational and commercial stability of the exchange, to all of the participants, at legally mandated meetings.

Its a shame if your experience at the local IXP has been less positive, perhaps look at the procedures and policies of a more sound exchange and encourage your local IXP to be run along those lines.

Andy

14

Sure, but we're talking about settlement-free peering. He's only expecting to be able to reach his peer's subnets and perhaps those of his peer's customers. If he peers with ASx in two locations, he does have redundant connections to ASx's tiny corner of the internet.

adam.

15

Paul Stewart wrote:

We have multiple transit providers today and are already present on a couple
of smaller peering exchanges with an open peering policy... our experience
with them has been very positive.

As an IX operator I'm glad to hear it :slight_smile:

The redundancy perspective is that you now have more paths to the same AS -
and an assumption that the peering route will always be best (I know that's
not always true).

Something to remember is that you are a network *operator* not a network
*purchaser*. If the peering route isn't working for you, pick up the phone and
talk to your peering partner. The whole point of being a network operator is
that you control who you connect with and take an active hand in fixing
problems! As others have stated, rich interconnection gives you greater
abilities in this area.

We of course have enough transit in case of a peering outage - would never
"put all our eggs into one basket" that it sounds like some others are

That attitude is quite 'old-school' - the idea that you can back up your
peering with transit often does not ring true in practice. You have less
visibility into your transit providers network than into your IXes networks,
and what information you do have is clouded by commercial concerns (read: sales
bullshit).

The traffic has to go somewhere, and if everyone in a metro area tries to send
to their transits it will just result in congestion within those networks -
even more likely when you consider the typical way their are built with ports
tiered off at layer2 from routers; traffic in the same metro area is likely to
simply hairpin up/down the router uplink.

Traffic between major transits within a metro area is also subject to
complicated commercial considerations which might mean the connectivity via
that route isn't so great.

also, we are looking at a number of them in various parts of the world
currently which adds another level of redundancy per say....

Many metro areas have more than one IX fabric often with considerable numbers
of operators on both. At LONAP in London we have members with big ports
expressly for backing up their private interconnects as well as to back up
sessions at other IXes.

In (primarily) Europe, the Euro-ix website has some useful resources to help
people select IXes: e.g https://www.euro-ix.net/member/m/peeringmatrix

Will

16

HRH Sven Olaf Prinz von CyberBunker-Kamphuis MP wrote:

internet exchanges are not per-se "redundant"
they basically are a switch which actually, because of the many connected
parties, most of which do not have enough PAID transit to cover any
outages on it, causes more problems than they are good for.
(the amsix with their many outages and connected parties that rely
primarliy on it's functionality is a prime example here)
  internet exchanges usually are some sort of hobby computer club, you
cannot rely on them to actually -work-, but when they do work that's
"nice" (always make sure you have enough paid capacity to cover for it
when they do not work however!)

peering on only one of them therefore does not make your network more
reliable in any way (it becomes a different story when you connect to like
10 or so worldwide).

as for "peering" agreements, just implement an open peering policy
(doesn't nessesarily have to take place over an ix, also applies to pieces
of ethernet running from your network to others).

those basically are contracts that force anyone who has also signed one to
peer with your network, wether they like you or not (saves the trouble
when you are a content provider and others do not want to peer with you
because they provide content too and you are a competing party etc).
  

Dear me, that smells of extreme ignorance of the design and management of the major exchanges.

LINX and AMS-IX for example go to great lengths to make sure their exchanges have high availability. I've had far fewer issues with individual exchanges with 100s of members than I have with single transit providers. The LINX for example provides TWO fabrics, and encourage members to peer on both of them. My transit providers have a single network which they break from time to time. It's far harder for an IX to break anything as they're less involved in the whole process.

It is true, of course, that there are tiny badly-run exchanges run as a hobby, but just as it's best not to buy transit from a bargain-basement transit provider, I wouldn't trust any important traffic to one of the tiny exchanges. I'd say that LINX/AMS-IX are amongst the most reliable places you can pass your traffic.

Since you bring up the "PAID" issue, as if to suggest that people who peer are cheap and don't care about their traffic, most organisations who peer do so to *improve* the performance of their networks. The cheaper route for me is not to buy a bunch of peering routers to manage 1000s of peering sessions, but I spend the extra cash to make the service I provide to my customers better. If you don't have the understanding or desire to provide the best service you can to your customers, perha1ps you'd like to become a politician?

Peering on one would make youre network more reliable if you have sufficiently burstable transit links. Only a fool would try to offload 180mbit of traffic via 100mbit of transit and 100mbit of peering. User stupidity isn't the fault of the exchanges and certainly don't diminish the viability of internet exchanges as a concept.

I think others have already rubbished your contracts nonsense, so I won't even bother.

adam.

17

Are you referring to his royal highness' network?
<http://www.cidr-report.org/cgi-bin/as-report?as=AS34109&v=4&view=2.0>

In any case, if you are really interested in achieving highly
reliable connectivity to the Internet, you need to start by
engineering it into your design. It's not just a matter of
choosing the right mix of peering and transit.

<http://en.wikipedia.org/wiki/Reliable_system_design>

--Michael Dillon

18

:-> "HRH" == HRH Sven Olaf Prinz von CyberBunker-Kamphuis MP <sven@cyberbunker.com> writes:

    > internet exchanges are not per-se "redundant"

depends on your concept of redundancy.

    > they basically are a switch which actually, because of the many
    > connected
    > parties, most of which do not have enough PAID transit to cover any
    > outages on it, causes more problems than they are good for.

depends who you peer with, and your comment on the IX being "a switch"
depends on which IX you connect to.

    > (the amsix with their many outages and connected parties that rely
    > primarliy on it's functionality is a prime example here)

How many of the outages at AMS-IX really affected you directly? or
weren't they rather limited to a bunch of your peers? And you know
that you can get multiple links to separate switches in different
location, don't you? Same goes for DE-CIX, LINX, the various Equinix,
PAIX etc...

    > internet exchanges usually are some sort of hobby computer club,
    > you

I think your choice of which internet exchanges to join has some
flaws.

    > cannot rely on them to actually -work-, but when they do work that's
    > "nice" (always make sure you have enough paid capacity to cover for it
    > when they do not work however!)

no, always make sure you have N+1 redundancy with a particular peer in
dispersed locations. or N+2 if you can afford the capex.

    > peering on only one of them therefore does not make your network more
    > reliable in any way (it becomes a different story when you connect to like
    > 10 or so worldwide).

    > as for "peering" agreements, just implement an open peering policy
    > (doesn't nessesarily have to take place over an ix, also applies to pieces
    > of ethernet running from your network to others).

    > those basically are contracts that force anyone who has also signed one to
    > peer with your network, wether they like you or not (saves the trouble
    > when you are a content provider and others do not want to peer with you
    > because they provide content too and you are a competing party etc).

you will find that most peering contracts or agreement have nice
clauses to terminate the peering at some agreed notice, as well as a
whole host of clauses that give the peering manager the power to say
no if he feels so.

    > --
    > HRH Sven Olaf Prinz von CyberBunker-Kamphuis, MP.

    > Minister of Telecommunications, Republic CyberBunker.

ok, you're a troll and I bit...

19

(the amsix with their many outages and connected parties that rely
primarliy on it's functionality is a prime example here)

internet exchanges usually are some sort of hobby computer club, you
cannot rely on them to actually -work-, but when they do work that's
"nice" (always make sure you have enough paid capacity to cover for it
when they do not work however!)

certainly looks like over 500Gb/s of traffic across ams-ix. that's a
big 'sort of hobby computer club'. i wonder what all those hobbiests
are doing.

in all seriousness, the above post is ludicrous. ams-ix runs one of
the most reliable exchange platforms on the planet due to an
incredible investment in optical switches and duplicate hardware.
it's expensive to run that way but the results have been incredible.

none of that is actually on-target for the original question about the
*value* (other than cost savings) of peering.

so far there have been some good values articulated and there may be
more (reach, latency, diversity of path, diversity of capacity,
control, flexibility, options, price negotation) and some additional
costs have been mentioned (capex for peering routing, opex for the
peering itself + cross connects + switch fees + additional time spent
troubleshooting routing events).

are there others?

Confidential: Please be advised that the information contained in this
email message, including all attached documents or files, is privileged
and confidential and is intended only for the use of the individual or
individuals addressed. Any other use, dissemination, distribution or
copying of this communication is strictly prohibited.

i was not an individual addressed but the attached mail was sent to a
mailing list of 10k people. HRH Sven Olaf is in violation of his own
policy about dissemination, distribution or copying.

t.

20

Almost certainly.

But I'm sure the OP has a nice list to at least get him started of peering benefits. Interestingly, no one has mentioned the downside of peering. Just to play devil's advocate, allow me to mention some "cons" about peering: If you drop all peering and push traffic to transit providers, you can frequently get lower price per bit. Picking 2/3/4 transit providers and committing large amounts to them gives you flexibility, control, reliability, lowers your CapEx, and lowers your network complexity which can (should) lower your OpEx. There are others, but you get the point.

Just be sure to consider everything when deciding whether to peer.