Peering & Transit - Public Peering

Summary

Peering is the direct exchange of mutual customer routing information and traffic between independent autonomous systems (networks). This can save both parties money on global transit costs and improve performance and resilience.

Peering does however require a lot of effort, and for diverse reasons not everybody will want to peer with you. A realistic target after one year is to obtain a 50% coverage of your Dutch traffic. The other 50% of your local traffic (and your International traffic) will then still be handled via transit.
As an alternative for global transit you can consider the NL Routing service of Open Peering which offers practically 100% local coverage.

Open Peering can negotiate, implement and maintain peering with the top-25 peers for you on your behalf at a tariff of € 3.000 initially and € 300 per month. Further Open Peering offers a FREE MultiLateral Peering (MLPA) service as an efficient solution for peering with the smaller parties.

Description

Advantages of peering

Delivering traffic via peering is much cheaper then delivering it via global transit;

  • A direct peering with an other party guarantees a very short and fast connection to that party and its customers: faster then otherwise possible;
  • Peering provides resilience as peers can be reached both via regular global transit and the peering. These paths back each other up;
  • Peering provides increased control the Internet traffic, reducing dependence on global transit providers;
  • Potential bottlenecks in provider or Internet infrastructure can be bypassed via "direct" peering path;
  • Peering can have a marketing value for a network, upgrading its status towards it customers;
  • For very large companies peering might be useful or even required because of government regulations to avoid the appearance of unfair use of their position as a substantial market power.

Limitations of peering

Although peering has very substantial advantages, it has some limitations and disadvantages that need to be taken into account to prevent disappointments:

  • A peering session with a party can only be set up after a peering agreement with that party has been established;
  • Each individual peer generally represents only a very small fraction of your total traffic. So to reach a traffic volume that is efficient (and saves cost) a very large amount of peerings has to be established;
  • Negotiating peering with a lot of potential peers is a lot of work, generally about one year in turn-around time, and a month in man hours;
  • For larger potential peers, peering with you is generally not very important. Such networks often do not respond to peering requests or process them very slowly. Especially with frequent takeovers and personnel changes, the response-rate of larger parties is generally very bad;
  • Generally a very small group of potential peers, the larger ones, represent a substantial part of your traffic. After that group the added value of each individual extra peer rapidly decreases;
  • Larger parties generally tend to have strict peering policies, e.g. demand a minimum mutual traffic level, equal in:out ratio of traffic, peering at multiple geographically dispersed locations (e.g. countries or continents), etc.
  • As a general rule of thumb, the larger and more interesting parties get, the more strict their peering policy becomes;
  • So even after a lot of effort to negotiate peering, after the first year it is very probable you still will not have peering with some of the peers that represent your largest traffic volume;
  • Peering requires continuous effort in maintenance.

Peering is limited to the respective networks themselves and their customers

Peering the exchange of routes and traffic limited to both networks and their respective customers. Peering does not include providing the other party with transit routing to non-customer networks or the rest of the Internet.

Peering is free

In the strict definition of the word, peering is settlement free: neither party pays they other party for traffic exchanged. If there is a financial settlement (one parties pays the other), this is called payed peering.

The difference between Bilateral and Multilateral peering

Bilateral peering means a peering relation between two parties, one-on-one, which is the most common form of peering.

Alternatively Multilateral peering is a peering relation between a group of networks, documented in a MultiLateral Peering Agreement or MLPA, in which participants agree to peer with all current and future other MLPA participants under the same conditions.

Advantages of Multilateral peering

With Multilateral Peering networks save a lot of time in peering negotiation, as with one agreement a network can arrange peering with a group of peers at once instead of negotiating with each of these parties one by one. Further the conditions of the agreement are the same for all peers and require a legal and technical review only once.

Generally the MLPA is also technically supported by a MLPA facilitator in such a way that it requires only a single technical configuration action to implement peering for all current and future peers, and not an action per individual peer. This saves a lot of time and work and guarantees the peering is immediately active when a new participant joins the MLPA.

Public versus private peering

Peering with an other party can be done over a dedicated point-to-point network cable between between the BGP4 routers of each party. This is called private peering (sometimes also called private interconnect in analogy to peering of telephone networks).

Alternatively peering can be implemented over the public infrastructure of an Internet Exchange. Then it is called public peering.

Why very large global networks are very selective in peering

The advantages of peering sounds evident, that on first glance it is difficult to understand why any party would not want to peer: each party saves money on transit costs and it improves network performance.

Very large networks however do so much peering with other large networks all over the world that they do not, or hardly, buy transit anymore. It is very probable they already peer with your upstream transit provider(s), so their traffic from/to you is already free for them. The cost saving is not really an argument for them anymore. And if they have enough peering capacity to your upstream(s), then there is no performance bottleneck on their side either.

Yes, you pay your upstream provider high transit costs for the traffic to/from that large network, and maintaining a lot of transit capacity can be very expensive for you. But that is your problem, not the problem of the large network.

Actually, if large networks would peer with you, they might even loose their own peerings with the other large networks (your upstream providers), as those other networks would loose transit income and not be happy.

For reasons like this the very large networks generally set very strict requirements for peering. They only want to peer if you are 'equal', in terms of traffic volume, in/out balancing, geographical network size, customer base, etc. Therefor peering with the largest networks is virtually impossible if you are not a very large network yourself as well.

Why large local and regional networks are also very selective in peering

For networks that are not very large on a global scale but are very large locally or regionally, like former national telephone companies, the same arguments count as for large global networks apply, on a smaller scale.

Such networks generally do have a strict peering policy, in order to protect what they consider their "home market": they have such a large local or regional market share that smaller local networks will pay for transit to them if they cannot access them via peering.

As such large local and regional networks do generally peer with each other, they also protect each others home markets and turnovers in this way: either you have to buy transit from one of them, or alternatively it helps them getting peering with the largest global networks.

Why even large DSL and cable access providers are selective in peering

Generally DSL and cable access providers are owned by a former national telephone company, and are forced to follow the same policy as their mother companies. Therefor they generally have strict peering policies.

Further they are generally peering for years and years already and have over time and historically build up a very good local coverage via peering. Therefor for them each individual new peer adds only a small amount of peering traffic and is not so important. Actually the chance is high that they would already receive the traffic of potential new peers via some other existing peer already without cost.

Each new peer does however add work in negotiation and maintenance, and new peers tend to make more mistakes then existing peers, potentially causing network and routing problems. Therefor peering with DSL and cable providers, although generally very interesting traffic-wise, is generally very difficult to obtain for new networks.

Legal implementation of peering

When parties agree to bilaterally peer with each other, this can be documented in a formal and signed peering agreement. Such an agreement formalizes technical best practices, the settlement free character of the traffic exchange, the duration of the agreement, the fact that it can generally be canceled at any time, etc.

In most cases however, bilateral peering is simply agreed upon via email and then technically implemented without a formal and signed agreement.

Multilateral peering is always documented in a formal and standard peering agreement: the MLPA. There is generally one party which functions as co-signer for all MLPA's and keeps a registry of all organizations which have signed the MLPA.

Technical implementation of peering

Bilateral peering is implemented by setting up a BGP session between routers of both parties over which routes of each network (autonomous system) and their respective customers are announced based on which resulting traffic is exchanged between the networks. Bilateral peering can be done both in public peering (over an Internet Exchange) and private peering (over a direct cable between the parties) form.

The technical implementation of multilateral peering can theoretically be a bilateral BGP session between two participants, just as with bilateral peering. In that case the role of the MLPA is limited to the legal agreement.

Generally the technical implementation of an MLPA is however done by setting up a BGP session with a an MLPA facilitator over an Internet Exchange. The facilitator will then redistribute all routes of all participants to all other routes. Payload traffic is generally exchanged directly between MLPA participants without the router of the facilitator in between.

How much manpower does bilateral peering require and what does it bring

What Effort & Result *
#potential peers 350
Turn-around time +/- 1 year
Man hours initial +/- 160 hours
(1 month)
Man hours maintenance +/- 2 days/month
(4 hours/week)
Success rate +/- 60%
(+/- 210 peers)
Dutch coverage +/- 50%

* All numbers are estimated averages based on experience

Documentation on peering

More info about peering can be found on Wikipedia.

Pricing

Full bilateral peering

Generally networks implement and maintain bilateral peering themselves. However, Open Peering does now offer the option to outsource this task, Open Peering is very experienced, and can generally arrange more peerings for your network and faster, without consuming substantial manpower:

Open Peering
Bilateral Peering - Full
Implementation and maintenance up to 350 potential peers
Initial Max 120 hours € 9.000,-
Monthly recurring Max 8 hours € 600,-

Top-25 bilateral peering

Considering that generally 20% of your peers represent 80% of your traffic, Open Peering can alternatively also arrange peering for you based on a selection of the top-25 largest networks on an Internet Exchange. This reduces the required manpower initially and in maintenance substantially, and inherently reduces the cost, without sacrificing substantial potential peering traffic:

Open Peering
Bilateral Peering - Top-25
Implementation and maintenance of core peering
Initial Max 40 hours € 3.000,-
Monthly recurring Max 4 hours € 300,-

Multilateral peering

Open Peering offers an MLPA Registry based on this standard MLPA Agreement with almost 100 MLPA Participants already.

Open Peering also offers an implementation of the MLPA as facilitator, on and between the AMS-IX, NL-ix and GN-IX Internet Exchanges. This means that parties which have signed the above MLPA get routing to all other signers on any of the above Internet Exchanges.

Open Peering
MLPA Peering
  MLPA Registry MLPA Routing
Initial FREE FREE
Monthly recurring FREE FREE

This MLPA routing service is NOT the same as the commercial NL Routing and Global Transit services Open Peering also offers. The MLPA provides +/- 5% Dutch coverage whereas NL and Global Transit offer up to 100% of local Dutch coverage.

Conditions

Bilateral Peering

Open Peering cannot make any guarantee regarding the percentage of Internet Exchange members will actually engage in a peering relation with your network under this service as it fully depends on the cooperation of a large amount of other parties. Open Peering will however will make any reasonable effort to get peerings arranged, within the agreed maximum number of hours.

MLPA Peering

Both the MLPA Registry and Routing services are free of charge and not for profit services of Open Peering, provided on a time-permitting basis without 24*7 support and theoretically be terminated at any point in time.