For our first-ever spotlight, we’re joined by Senior Analyst Paul Brodsky. Paul speaks with us about how - quite literally - the information we share traverses the internet.
Expanding on his popular TeleGeography Takeaways explainer on IP transit and peering, we talk about the mechanics of IP transit and peering, as well as geography's role in where your data travels before it reaches a final destination.
This interview has been condensed and edited.
JM: Paul, we know that you have done several series for us about main TeleGeography takeaways – big picture stuff. You were recently involved in a video that was all about IP transit and peering. You broke it down, explained the ins and outs. Can you give us a walkthrough of what that was about and some of the details you explained in this recent piece?
PB: What I find interesting are the commercial arrangements by which these network operators actually deal with each other in forwarding their data.
It takes two basic forms. The first, as I discussed, is called IP transit, which is a – from a commercial standpoint – a very simple, straightforward, metered service where the customer moves his IP traffic upstream to the IP transit provider who in turn is responsible for the ultimate routing of the traffic to the rest of the internet.
Now, again, even the biggest IP transit providers can not necessarily move that traffic to every end point on the internet. So the IP transit provider in turn needs to have arrangements with other network operators to move the traffic to its ultimate destination.
JM: Does location make a difference? How does the geography of the internet – something we talk about here a lot – how does that come into play?
PB: Yes. And it varies where you are in the world, also. In some parts of the world there are initiatives to get local ISPs to peer with each other. Some of the developing part of the world. In Africa, there is this nascent drive to get local ISPs to meet each other, locally, in Africa to exchange traffic.
Up until very recently – and still very much today – believe it or not, if you have a network operator in Kenya and a network operator in Tanzania, they’ll exchange traffic in London. And it will go through an intermediate ISP in order to do that. So the Kenyan ISP will actually purchase IP transit in London. The Tanzanian ISP will purchase transit in London. And that’s where those packets will be routed.
JM: It’s cheaper, right? Everyone does it, but then you have to wait forever.
PB: It’s cheaper! Right. So there are these initiatives now – and I’m just using Africa as an example – to try and encourage direct interconnection of these networks in Africa. To avoid this tromboning, as we say, of data from Africa, to Europe, back to Africa again.
JM: What is the decision process like for someone managing transit and peering arrangements for a company? What comes to mind when selecting one or the other, combining these methods? What are the questions being asked or the key considerations?
PB: There’s no one right way for a network operator to route traffic.
These are all business decisions that these ISPs make. One way that ISPs may choose to have their traffic routed is to start with more than one IP transit provider to begin with. Because you never want to rely just on one IP transit provider. If that provider goes down, then you’re down.
So, typically, any self-respecting ISP will purchase IP transit from multiple transit providers for redundancy’s sake.
And then as they study the way their traffic moves, they may realize that a fairly healthy chunk of their traffic is being routed to only a small handful of network operators. And they may decide, then, to peer with those network operators in order to reduce their reliance on IP transit, and thereby reduce their IP transit cost.
Again, with the caveat that peering in and of itself is not cost-free. It’s often settlement-free. But there are costs incurred to peer. And so, taking all these costs into consideration, this is how network operators choose to construct their network interconnection arrangements.
There are some – in some of the more developing parts of the world, there are these seemingly illogical arrangements where network operators that might be right next to each other don’t meet each other locally. They go to some far off place to exchange their traffic via IP transit. And that’s starting to change. Although there is some institutional inertia that needs to be overcome in order to get some of these network operators to agree to simply peer with each other locally.