One of the confusing aspects of Internet operation is the difference between the types of providers and the types of peering. There are three primary types of peering, and three primary types of services service providers actually provide. The figure below illustrates the three different kinds of peering.
One provider can agree to provide transit for another provider. In this case, one provider is agreeing to transit (or carry) traffic from the other provider to its peers. In this type of relationship, one provider pays the other based (called settlement) based on the amount of traffic carried (not the link size!). While transit peering relationships are normally used by providers closer to the edge of the Internet to reach closer to the core, and is used by providers selling customers Internet access (or other services). Transit peering arrangements normally assume both providers are actually transiting traffic among multiple autonomous systems.
The second most common type of peering is provider to customer. While this might seem similar to transit peering, it is actually a completely different thing altogether. In provider to customer peering, the customer generally pays the provider based on the connection speed (although there are many situations, such as mobile providers, where customers are charged based on utilization, like in a transit peering relationship). It’s also assumed that customers are “valley free,” which means they are not transiting traffic between other autonomous systems; traffic can be sourced or sink’d in a customer network, but should never pass through the customer network from one provider to another.
Finally, there is settlement free peering. This is similar to transit peering — except it assumes the traffic flow in both directions is about the same (in terms of actual usage). There is no payment (settlement) between providers who are in a settlement free peering relationship. Settlement free peering is normally entered into by two providers who both believe they are providing better service for their customers through the peering arrangement, while not giving any competitive advantage to their peer.
This might seem to be very clear — every type of peering has it’s own definition, and it’s own set of rules. In the real world, things are quite a bit messier than this. As the traffic flow between two large providers can change from day to day, there may be some days where no settlement is charged, others where one provider is charged, and still others where the other provider is charged. Is this a settlement free or transit peering relationship? It all depends on the day… Or what about large scale content providers — do they always have provider/customer relationships with other providers? No, not really — in fact, some content providers have enough leverage to insist on settlement free peering in most places, or even to charge for caching and other services to optimize a peer’s network.
And what about a large enterprise network that actually carries traffic from vendors and guests through their network (autonomous system) to an upstream provider? Is this enterprise a transit provider, a customer — or a little of both? As you dive into the routing ecosystem, you find a lot of situations where the lines are blurred, and answers aren’t easy to find. Don’t let it dissuade you, though; start with the easy cases, and let the more difficult cases clear up over time.
Next time I’ll go into the services offered, which are often confused with the type of provider — and we’ll look at why it’s really hard to classify a provider based on their “primary” services.