If you’ve worked at an ISP or even just closely with them you’ve probably hearing the term peering quite a bit. Peering is essentially a reciprocal agreement to provide access to networks between two providers. Provider A agrees to allow Provider B to send traffic over and through their network in exchange for the same access in the other direction. Sounds easy, right? On a technical level it is pretty easy. You simply set up a BGP session with the partner provider and make sure all the settings match and you’ve got things rolling.
The technical part isn’t usually where peering gets complicated. Instead it’s almost always related to the business side of things. The policy and negations that have to happen for a good peering agreement take way more time that hammering out some BGP configuration stanzas. The amount of traffic to be sent, the latency requirements, and even the cost of the agreement are all things that have to be figured out before the first hello packet can be exchanged. This agreement is always up for negotiation too, since the traffic patterns can change before you realize it and put you at a disadvantage.
Peerless Data Collection
If you want to get the most out of your peering arrangement you need to know what’s going on. You need to have statistics about the key points of your agreement. You need to know if you’re holding up your end of the bargain as well as the company you’re working with. If you walk into a peering negotiation without the right data you’re going to be working from a disadvantage right away.
For example, did your partner company take all of the traffic they agreed to accept in the peering contract? Or did they have issues that forced you to send the traffic along a different route? Were you forced to send that traffic across a different route that had a higher cost? Did your users complain about network speed because of congestion outside of your control? If you can’t put your fingers on the answers to these questions quickly you’re going to find yourself with lots of angry users and customers not to mention peering partners that want answers from you as well.
Recently I had a chance to listen to a great presentation from Kentik during Networking Field Day: Service Provider. Nina Bargisten laid out some of the challenges that Kentik customers face with peering arrangements and how Kentik is helping to solve them:
One of the points that Nina discusses is that capacity planning is a huge undertaking for ISPs. With the supply chain issues that we’re currently facing in 2022 it’s not easy to order equipment to alleviate congestion problems. Even under somewhat normal circumstances it’s not likely that an ISP is going to go out and order a lot of new hardware just to deal with congestion. They might change some polices to route traffic in different directions but ultimately the decision has to be made about how to get customer packets through and out of their network to the ultimate destination.
Peering agreements can help with congestion. Adding more exit points to your network means some flows can exit through a different provider and either get to their destination faster or prevent a larger connection from being overwhelmed and congested. It’s not unlike having multiple options to use to arrive a destination when driving. Some streets are better for smaller amounts of traffic compared to larger highways and interstates that provide high-speed travel.
As mentioned above, it’s critical that you have data on your traffic and its performance. Are you sending everything through one route? Are you peering with providers that are getting less than half of the traffic load they agreed to take? These are all questions you have to ask to create a capacity plan. If you’re hearing complaints about congestion but you see that only two of your outbound connections are running a full capacity while the rest are sitting idle then you don’t have a congestion issue as much as you have a configuration problem to solve.
Kentik’s solution allows you to see what’s going on and help you make better decisions about the routes that traffic should be taking. As demonstrated above, their dashboard collects data from your network as well as many others and can tell you when you need to be configuring polices to send traffic to low volume peers instead of relying on congested links. It will also help you see trends for when links become congested and allow you to set thresholds to divide your traffic appropriately before it becomes an issue.
There’s a lot more info in the video above to help you with your capacity planning and peering negotiations. It all comes down to a simple maxim: Information is key. You can’t solve these puzzles without knowing what you have and what you need. If you’re just going to keep throwing peering agreements at a problem until it goes away you’re going to fail. You won’t solve your real issues by just adding another connection that never gets used. Instead you can use Kentik’s platform to provide the kinds of insights that will help you create value for your customers and save money at the same time.
Service providers think about traffic differently than enterprise admins. They have to worry about it coming into the network and leaving again. Instead of worrying about a couple of links to the wider Internet they have to worry about dozens. If you think it’s hard keeping track of all that data for the enterprise you can just imagine how hard it is when you scale it up to the service provider level. Thankfully companies like Kentik are applying their expertise to provide actionable information to help you make the right choices and maybe even negotiate some better deals.