Business Model Premise: Operation of the Internet distribution is seen as strategic to the VSP:

a)End-user experience is mission-critical so outsourcing the end user experience to a transit provider or CDN puts the VSP at risk.

b)The VSP has visibility into what video are being released, which ones are likely to be hot and which ones don’t require special infrastructure adjustments.

c)Internet Video distribution is so new that the VSP prefers control. This is a strategic focus for the VSP: ensuring reliability, scalability, through the constant monitoring and evolution of the infrastructure to ensure the end user experiences during these early phases of Internet Video Distribution.

d)The traditional CDN may be ill-suited to distribute very large video objects, therefore we have to do it yourself.

One reviewer points out a few additional motivations for some companies to build their own CDN:

e)Content producers (Movie Studios for example) may want to bypass distributors because they can, whereas they could not bypass distributors before the Internet.

f)Some profiles of web objects are more difficult for CDNs to handle well: large volume of accesses of large objects and small volumes of accesses of small objects don’t fit the CDN model as they rarely are served out of CDN edge cache.

g)Empire Building – some DIY CDNs are built not out of necessity but by staff need to build their own importance within the organization, aka build and operate the infrastructure ‘better’.

Definition: Peering is a business relationship whereby two entities reciprocally exchange access to each others networks.

Is Peering for Content?  Many of the largest Content Providers in the world are peering now. There are three primary motivations for peering. Peering provides the lowest latency path between two points. Peering bypasses the transit provider providing cost savings for both parties. Peering also provides greater control over the end user experience. Yahoo!, Microsoft, and Google for example have built out substantial networks and are peering at exchange points around the world. Yahoo! currently has over 640 peering sessions and a multiple- OC-192 (10Gbps) global backbone to distribute its content itself to peering and transit relationships.

Peering is not Free. There is a recurring flat monthly fee associated with public peering (switch port fees, collocation expenses, etc.) and private peering (cross connect or circuit fees, collocation, etc.). To compare peering against transit, one compares the unit cost of transit (in Mbps) against the unit cost of peering (in Mbps). Since peering costs are fixed monthly, the unit cost of peering varies based on the amount of traffic peered during the month.

The Peering Break Even Point (where the unit cost of transit equals the unit cost of Peering) is calculated using a lot of assumptions but in 2006 approaches 300Mbps. That is, the cost of peering is completely covered by the cost savings of peering at least 300Mbps of traffic that would otherwise traverse a metered transit service. Fortunately, as we will see, video traffic is so large that this breakeven point is met in even the light load model.

Con: Requires the Content Provider become a Network Operator. This means a content provider interested in peering needs a 24/7 NOC, networking expertise on staff, etc. a seasoned and talented Peering Coordinator is required to obtain peering. These requirements can be expensive and away from the core competence of a video service provider. On the other hand, some of these things may be already be in place, perhaps as required for some other operations activities.

We will make a number of simplifying assumptions for all 3 Transit/Peering models:

Three IX Deployments. Jeffrey Papen points out that three sites are sufficient to meet most geographic diversity peering prerequisites for the eyeball networks that would receive video traffic. More sites than that he argues would provide diminishing returns; the costs may exceed the incremental peers picked up.

No Backbone. These models assume that no backbone is used for the distributed VSP implementation based on Jim Gray (Microsoft) assertion that the most cost effective way of distributing large amounts of content is to UPS overnight disk drives. So, the servers and routers are shipped pre- configured with videos. It further assumes that updates (new releases) are infrequent and are distributed over a tunnel using a commodity transit services.

Each deployment independently can handle load. For simplicity, we will use three copies of the same equipment, ignoring efficiencies of distributing the load across all IXes. (If we assumed a uniform traffic load distribution across all IXes, we could for example chose fewer servers and smaller routers.) So Model 1A, 1B and 1C will each be deployed to three IXes across the U.S. resulting in three times the deployment cost.

Purchase transit, peer in each location with an open peering policy and expert Peering Coordinator(s) as part of the network operations group. Active expert peering evangelism is required given the following five hurdles:

1)BGP is not perceived as a content providers expertise, so ISPs may be skeptical of their network competence. Since they will be woken up in the middle of the night, the ISP Peering Coordinator may not be receptive to peering discussions with perceived novices.

2)Peering requests may be seen as a revenue opportunity by the ISP approached. Sales Revenue is preferred by the ISPs, particularly given the large amount of traffic the ISP is expected to receive from the VSP.

3)Personality Clashes have prevented peering from going forward where it otherwise is in both companies best interests. A skilled Peering Coordinator knows the venues, the personalities, the approaches that will work and already has a rolodex with the right contacts to initiate discussions.

4)The potential load is so large that the peer may need to rapidly upgrade peering infrastructure. Without revenues to support this activity, there may be reluctance for the ISP to peer.

5)The peer may see a business clash in peering if their company sells videos to their customers also. They may not want to encourage or enable a far away competitor that will use their bandwidth to compete with them (aka the Net Neutrality Debate).