The Three-Day Monty


DrPeering - 

Eyeball networks won’t peer with me. What levers exist to pressure them to peer?

-- Content Carl


Carl -

As a curator of Internet Peering tactics, and I am happy to share a new tactic in the wild, contributed anonymously as usual. This tactic seeks to obtain peering by making it painful not to.

This tactic is called the “Three-Day Monte.”

Like the Traffic Manipulation tactic, this maneuver leverages the fact that the content side controls the path of the content distribution. Like the “Gaming the 95th percentile” tactic, this tactic leverages the fact that the peak 36 hours of monthly traffic is free for a transit customer. Consider the diagram below.

Assume that Content Provider (A) wants to peer with Access Provider (B), but the B might get that A traffic for free across one of its current peers (Peer #1 or Peer #2). They already get your traffic for free. Or, B might get your peak A traffic for free because the short duration spikes in traffic flow from A to B fits nicely into the current peak 36 hours of one of its transit providers. In either case, there is no pain to B that peering with A would address. Yet.

So A rotates its outbound traffic flow across all upstreams for at least three days at a time to maximize the likelihood that the three days of traffic arrive on a path that the access network has to pay for. Now peering with A would have a direct financial benefit.

Once peering is established, traffic to B will naturally flow to the directly attached peer B, regardless of the transit provider manipulation that A does. The traffic to B will from here forward prefer the direct path regardless of the paths the rest of the traffic takes during the continued rotation. 

Pretty clever.