Dear ask@DrPeering.net ,

Q: I heard that peering can potentially save money and improve performance, but can peering actually increase revenue?

Sincerely,

Pierre Conè

---------------

Dear Pierre -

Thank you for that question. We focus so much time on the Peering Breakeven Point (the point where the cost of peering exactly equals the cost of transit) that we tend to overlook the other benefits of Peering, including incremental revenue flow.   The logic came from the field, from Dave Rand of AboveNet fame in particular, and it goes like this:

Internet Service Providers charge transit fees based on the amount of traffic delivered. Therefore, the more traffic delivered, the more revenue the ISP makes. 

So what hinders traffic volume?  

1. a)Packet loss. When a TCP traffic flow encounters packet loss, the TCP window size (amount of traffic in mid-air between the two parties) divides by two, and then slowly grows back up. *So every packet lost impacts revenue! *
   

2. b)Latency. The latency between endpoints affects how long it takes for the TCP window to open, so especially for short duration flows (like web pages and p2p), high latency has the effect of slowing traffic flow somewhat.
   

Peering, by definition, provides the *direct* exchange of traffic to each others customers. The resulting shorter path (generally) means lower latency, so the TCP window opens up faster, more traffic is exchanged, more revenue is generated.  More importantly, peers have some ability to shut down, or redirect around congested paths.  

By the way....DrPeering has searched for years for some way to quantify the incremental revenue made by peering. How might one go about quantifying this?

It is also interesting to note that the content companies that peer say "It is all about end-user experience." The #1 reason they gave for peering was to have greater control over routing, so that when they see packet loss (visible in the host server statistics), their network engineers can do something to address it. When a page loads slowly, the user goes to an alternative sites, decreasing their page views and corresponding ad revenue. This is one reason why Google, Yahoo!, Microsoft, Sony Online, Electronic Arts, etc. are "open" peers.

We'll talk more about "open" peers next time.

The final argument heard is that

1. a)peering may provide multi-homed customers with a shorter AS-Path for traffic,
   

2. b)this traffic will prefer the short path, therefore
   

3. c)peering increases revenue from multi-homed customers
   

DrPeering