Hello everyone and welcome to the next article in our series about the Internet. To recap, we introduced this series by examining the history of the Internet and the conceptual tenets that laid the groundwork for the massive network that is the Internet today. In this article, we will examine the three main types of Internet connections that are available and how they fit together to build the Internet. Now let get’s started.
CCNA Training – Resources (Intense)
We already established from the last post that the Internet is not owned by anyone or any government. It is the strategic connections of many autonomous networks that form the Internet, and the whole is larger than the sum of the individual parts. So how do you connect to the Internet?
A transit connection to the Internet is a connection used as a middleman between a source and its destination(s). So when we connect to our ISP from homes or offices, we are “hiring” our ISPs to “carry” traffic from us to wherever we want to get connected to.
Similarly, our ISP would also have transit connections to other larger ISPs to carry traffic from its autonomous system (including those of its clients) to other destinations they’re not directly connected to. This is illustrated in the diagrams below:
In the first diagram, source A has a transit connection to ISP C. Similarly, Destination B also has a transit connection to ISP C. In this setup, both A and B would pay some money to C for the transit service that it provides. It is also important to note that A and B are not particularly interested in communicating with C (hence the transit function) but only with each other.
The second diagram is similar to the first with the only difference being that C now has a transit connection with D, while B has a transit connection to D rather than C. In this setup, money exchanges hands in three locations:
* A pays C for its transit service.
* C pays D for its transit service.
* B pays D for its transit service.
What determines how much money is paid in each scenario? Like any other business, the amount of money paid is determined by fixed costs and variable costs. However in this case, the fixed costs (installation and equipment) are usually minimal when compared to the variable costs (bandwidth used). So the major determinant of cost paid is the amount of traffic sent through a transit.
These simple setups show connections between two endpoints through one or two transits. Once we introduce more connectivity requirements, we can easily see how the Internet would breakdown into economic chaos with transits all over the place and money exchanging hands between multiple providers and their clients. This is what is called the economy of the Internet.
In some cases, however, Autonomous Systems can decide that rather than paying each other for transit features, there is a mutual benefit in exchanging traffic for free between each other. This is new type of connection is called peerings. Let us now examine peerings in detail.
Considering the second diagram, what if D needs to send traffic to customers of C? It would not make sense for C and D to pay each other for mutual transit features. So instead of having transit agreements, they would have peering agreements. In a peering agreement, both parties are still autonomous but they agree to send traffic to each other’s customers without paying each other. So revenue is generated from their customers (A and B) as shown below;
Another scenario where peering agreements are good is when two organizations send a lot of traffic to each other (for example, Facebook and Google). In this case, it is wiser for them to peer with each other directly rather than send traffic through transits (and end up paying tons of money for the transit service). So all traffic sourced in FB destined for GOOG and vice versa is sent over a private link.
A = FB
B = GOOG
There are some considerations to make before peering is done:
* Size of the peer: Normally, networks of similar sizes peer with each other for mutual benefits.
* Locations of peer: This is closely linked with size. You would normally want to peer with a network that can provide similar location advantages. If an ISP only has presence on the East Coast, it is unlikely to peer with an ISP that has presence across the entire US because the larger peer would prefer to be a transit. In that case, they can enter a peering agreement for the East Coast and a transit agreement for the West Coast.
* Traffic engineering: We will discuss this in more detail later but essentially, organizations would prefer to send traffic across free links since it is cost efficient. Peerings must ensure that they do not get used as transit for networks that they might end up paying for.
There are two types of peerings: There is the private peering which has been described above. In this case, two organizations decide to enter a peering agreement and they connect to each other physically (yes, we need to have a cable plugged into a device at each end!) and peer over that physical link. However, with the rise of so many Autonomous Systems, full mesh private peerings in the same location have become economically unscalable due to high costs of equipment, installation and operations. So instead of having private peerings, we would have public peerings via Internet Exchanges.
Public peerings are achieved by connecting multiple peers to a shared fabric (usually a layer 2 access). These neutral locations where public peerings are held are called Internet exchanges, previously referred to as Network Access Points (NAP). I would spare you bits of history on these because I think they are boring and do not necessarily contribute any real knowledge . These peerings are usually operated and maintained by a non-commercial, neutral third party and are governed by peering agreements and templates. Large Internet exchanges usually have hundreds of members connected in a collocation facility.
Again, we should note that all members of an exchange are still autonomous and control how and whom they decide to peer and route with. Popular Internet exchanges include the London Internet Exchange, the Japan Internet Exchange, the New York Internet Exchange, the Los Angeles Internet Exchange, etc.
Large ISPs usually prefer to act as transits rather than have peering connections since it is more economically viable. And when they decide to peer, they would rather have one off private Internet peerings with equally large ISPs than connect to an exchange point. Eventually, the largest ISPs (also called Tier 1 ISPs) in a region peer with those in their region and other regions to ensure connectivity across the globe.
In this article, we explored Internet connections by examining the basic kinds of connections that exist between networks on the Internet today. We also saw how financial economics drive the kind of connections that is established between organizations on the Internet. I hope this has laid a good foundation for the next article which focuses on how the Internet Protocol is used to establish connections on the Internet.
As usual, if you have any questions or comments, please feel free to use the comment box below to share your thoughts. Thanks again for reading and I look forward to writing the next article in this series.