Like It or Not, Content Delivery Networks Will Pay Eyeball Networks Because Traffic is Highly Asymmetrical

March 06, 2014
By: Gary Kim

It might go without saying that reliance on the Internet for communications, entertainment, work, and learning has implications for policy and business relationships within the ecosystem.

Network interconnection is one example, but a wider set of issues has arisen, around the traditional “any to any” communications value of the older Internet, application blocking, desirable business models for app providers or Internet service providers.

The differences between traditional telecom service provider interconnection, which has been highly regulated, and interconnection of Internet Protocol networks and domains, which historically has been unregulated, provide one example.

In the telecom interconnection environment, it always is assumed that carriers will compensate each other for terminating traffic delivered by another connected network, on the logical assumption that terminating that traffic has direct costs.

IP networks historically have used different rules, based on settlement-free “peering.” That worked fine when traffic flows were roughly symmetrical. The system breaks down when traffic flows are asymmetrical. Then small IP networks pay transit fees to larger IP networks to interconnect.

Some think IP networks should not have business relationships more like carrier interconnection than IP network peering.

But highly asymmetrical traffic flows characteristic of content delivery networks and video apps already is forcing a change, as network costs matter just as much for interconnecting IP networks as for interconnecting “telco” networks.

Content distribution, especially of video, already drives the bulk of Internet bandwidth demand.  In fact, video will drive 61 percent of bandwidth demand by about 2015, Cisco (News - Alert) has predicted. And most of that demand will be highly asymmetrical.

That will have implications for interconnection policy and business practices. Simply, content delivery networks always will represent high traffic loads for a terminating IP network, and therefore will drive a disproportional share of terminating network capital and operating cost.

Those costs will have to be recovered, by the terminating network, from the originating network.

Some will work about the market impact of payments to terminating networks, by originating networks. But those changes will come; they have to. Content consumption creates asymmetrical network demand. That demand creates direct cost, borne by the terminating IP networks. Originating networks will have to pay. 




Edited by Cassandra Tucker