The Implications of Mediated Peering & Transit (Pt. II)

In my last post, I reviewed what it means to offer "diamond, truck, and best-effort lanes" for service levels on the network. The need for diamond and truck lanes today is primarily an economic one. Savvy entrepreneurs and technologists have effectively figured out how to obtain bandwidth that is nearly "free" through the use of peer-to-peer (P2P). P2P places the economic burden on the last-mile broadband providers by allowing end-consumers to "share" their bandwidth with one another without any compensation to the ISP. In effect, this free bandwidth has created the vast majority of congestion on today's broadband networks. In some circles, it is estimated that BitTorrent alone accounts for 60% of the broadband bandwidth consumed.



This situation would be analogous to giving drivers in already congested regions of the highway "free gas". The impact of allowing people to drive for free would only increase the already congested highway system. P2P, because of its very economic nature acts as yet another application that will fill up the pipes to consume all available bandwidth.

Bandwidth just like the memory in a computer are both scarce resources. They also both have different associated quality characteristics. High speed "cached" memory is much more valuable to some applications than ordinary RAM. Similarly, bandwidth that loses very few packets and offers guaranteed minimum service levels is far more valuable than just ordinary "best-effort" bandwidth like we have on most broadband connections in the Internet today.

If diamond lanes are offered without the corresponding truck lanes, the P2P applications that can easily obtain that "free" best-effort bandwidth will simply use up most of the capacity. Consumers and broadband providers will be left holding the bag. Unless we relegate the "free use" of the best-effort bandwidth, that is the P2P traffic, to the truck lane, then consumers are going to have an unsatisfactory experience surfing the web. This type of challenge is already occurring with the broadband providers trying to figure out how to mitigate use of all that P2P traffic. Today, unfortunately, they're primarily left with one and only one solution: keep expanding those highways... Free loading P2P distribution systems have no economic incentive to slow down or use less bandwidth. They'll just keep going until every last drop that can be consumed will be consumed by their applications distributing pirated content.

Peering Amongst Equals

Once our initial InterStream Tier 1 networks interconnect under the Mediated Bandwidth Agreement (MBA) we're developing, a new model will emerge. At a minimum, the ISPs who participate in this agreement must offer to interconnect their premium video service offerings with one another. In other words, if one ISP connects with another under the MBA, they must each agree to give the other premium access to one another's network. This is the reciprocal nature of the agreement. In addition, there is a transitive nature to the agreement. If an ISP peers with a another different ISP, they must agree to extend the agreement to all of the other peers. As such, each of the ISPs in the network naturally extends their reach for premium services as additional peers come into the association. IP transit services are similarly extended by the peers just as we currently have with our IPv4 and IPv6 "Tier 1" networks today. Mediated peering and transit will have a dramatic impact on the evolution of the wholesale Internet.


The implications of this agreement could be quite profound. Tier 1 ISPs who offer premium last-mile access will be able to gain an early mover advantage in offering a new premium wholesale service. Other wholesale providers will be encouraged to follow suit and peer with these new providers. InterStream expects not only a new Internet bandwidth commodity to emerge but potentially a new set of "core" wholesale ISPs may emerge as well. Because video consumes such an enormous amount of bandwidth, new traffic exchange ratios may evolve out of the existing peering infrastructure already in place.

Pilot Implementation Objectives

Of course all of this is speculative until some hard data emerges from our pilot implementation. One of the key objectives of the pilot is to determine what new types of traffic ratios should be expected from high quality Internet video distribution. In addition, InterStream expects to learn what mediation policies must be set at the completion of the pilot. For example, with the "new math" I described in my previous post, it will now be possible to guarantee "best-effort" service levels to individual broadband consumers. In other words, broadband providers can objectively measure when they should be shunting the P2P traffic off to the "truck lane" so that their consumers do not experience any degradation in service. InterStream will demonstrate both the math and business model through the pilot.


What About Private Interconnect Agreements?

Mediating peering and transit does not imply that ISPs must follow an "open peering" policy. It has no implications on moving beyond the U.S. Title I agreements that most of the Internet runs on today, either. However, the "diamond lane" connectivity supported by the MBA does imply that service must be transparently offered. In other words, the terms of peering and transit contracts can be just as tightly specified under existing non-disclosure agreements while allowing advertisement of the service to all peered ISPs and transit providers. The MBA will require both last-mile ISPs and backbone providers specify that they're offering InterStream service by meeting required quality standards. By doing so users and customers of the media grid can be assured that they're getting the service they require. Non-paricipating ISPs won't get the benefit of either the service mark or early mover advantages.

If you have interest in more information or just want to give us your thoughts and feedback, please feel free to email us at info [at] interstream [dot] com.

Jeff Turner


Post new comment