The contents of this legacy page are no longer maintained nor supported, and are made available only for historical purposes.

Economics of Internet Interconnection

This page summarizes our project studying the economics of Internet interconnection. For more details please see the project's funding page.

Goals of the project

The high-level objective of this research is to create a scientific basis for modeling Internet interdomain interconnection and dynamics. Specifically, we aim to understand the structure and dynamics of the Internet ecosystem from an economic perspective, capturing relevant interactions between network business relations, internetwork topology, routing policies, and resulting interdomain traffic flow. Despite much recent interest in the economic aspects of the Internet, such as network interconnection (peering), pricing, performance, and the profitability of various network types, two historical developments contribute to a persistent disconnect between economic models and actual operational practices on the Internet. First, the Internet became too complex -- in traffic dynamics, topology, and economics -- for currently available analytical tools to allow realistic modeling. Second, the data needed to parameterize more realistic models -- interdomain traffic characteristics, routing and peering policies and pricing/cost structures -- has simply not been available. The problem is fundamental, and familiar: simple models are not valid, and complex models cannot be validated.

We propose a computational approach that promises progress in in both dimensions: creating more powerful, empirically parameterized computational tools, and enabling broader validation than previously possible. We will use measurements of interdomain traffic, topology dynamics, routing policies and peering practices as input to our detailed model of AS interconnection, and compute the equilibrium -- a state where no network has the incentive to change its connectivity. To validate our model, we will verify that it can reproduce known macroscopic properties of the Internet AS topology as well as known trends in Internet evolution, based on publically available financial and topological data. We will then use our model to study various interconnection practices, the stability and dynamics of interdomain links, and economic properties of the resulting equilibrium.

Specifically, we aim to measure the following properties of the interdomain Internet.

  • Interdomain traffic characteristics directly measured from different vantage points on the Internet.
  • Structural characteristics of the Internet's interdomain topology, its evolution over time, and the the economic implications of these properties.
  • Interdomain routing policies used by networks and the economic incentives behind those policies.
  • Peering policies used by different network types as inferred from publicly available information such as peeringDB and Internet Exchange Points (IXPs).

The data by itself has the potential to yield important insights, but we will use this data to parameterize and validate our model of Internet interconnection evolution and dynamics. Each iteration of the model executes a provider and peer selection strategy for each network in turn, until it reaches a equilibrium state, in which no network has the incentive to unilaterally change its connectivity. Why is it important to study equilibria of an Internet ecosystem that is always in flux? We believe that given a certain set of external conditions (interdomain traffic patterns, pricing/cost structures, routing/peering policies), studying the resulting equilibrium can give us insights into the "best" that networks can do, as well as the properties of the global Internet with respect to topology, traffic flow and economics. The equilibrium we compute will be specific to the set of these external conditions. By studying equilibria that result from different parameterizations, we can answer a variety of what-if questions about the evolution of the Internet, including:

  • The effect of changing traffic patterns on the economics, traffic flow, and topology of the resulting internetwork. For example, we can study the effect of the rise of large content providers that generate a significant portion of Internet traffic, or the rise of a popular high-volume P2P application.
  • The effect of changing price/cost structures, such as cheap peering and transit.
  • The effect of the increasing popularity of Internet Exchange Points (IXPs), which facilitate easy and inexpensive peering.
  • The effect of the increased use of paid-peering, where routing decisions are similar to settlement-free peering, but one network pays the other.

Data collection

We will begin with raw data from the following data sources to capture economically-relevant information about interdomain topology, traffic, routing and peering policies.

Data type Description Source
interdomain traffic We will extract the source and destination ASes from netflow records collected from various vantage points, to measure interdomain traffic patterns. Georgia Tech, UCSD, Internet2
Interdomain topology BGP data from Routeviews and RIPE can be used to reconstruct interdomain topology snapshots Routeviews, RIPE
Interdomain topology Traceroute data from Ark can be used to infer AS-level connectivity. CAIDA's Ark infrastructure
AS relationships The business type of interdomain links inferred using AS-relationship inference algorithms. CAIDA's AS-relationship project
Routing policies Routing policies used by ISPs as inferred using AS topology and AS relationship data AS topology, AS relationship data
Financial data Information about company revenues, profits, and stock prices. SEC filings, public financial data

Additional traffic data needed

Interdomain traffic characteristics are an important input to our model, as transit payments, peering costs and operational costs all depend on aggregate traffic volume. Unfortunately, there is little knowledge of the global Internet interdomain traffic matrix, even a snapshot much less its dynamics. Interdomain traffic statistics collected from a transit provider network will allow us to infer at least qualitative properties of the interdomain traffic matrix, such as the distribution of traffic sent (received) by a network to (from) every other network. Is it Uniform, Zipf, or Pareto? How does it differ by network, and change over time? We can also use measured traffic characteristics to evaluate "what-if" scenarios that arise from changes to interdomain traffic characteristics. Traffic estimates may even help us to improve our AS-relationship algorithms, based on observed correlations between inter-AS peering likelihood and inter-AS traffic volume ratios. If you are in a position to share data under appropriate privacy and acceptable use agreements, please email info@caida.org.

Project Deliverables

  • We will release periodic snapshots of all data we collect, subject to private data-sharing agreements, to enable the study of the evolution of interdomain topology, traffic patterns, routing policies, peering policies, and the financial performance of ISPs.
  • We will release the software implementation of our model to enable other researchers to investigate different parameterizations and what-if scenarios.
  • We will disseminate --- through publications at conferences, journals and network operator venues such as NANOG --- insights obtained from the measurements and by applying our model to various what-if scenarios.

Publications resulting from this project

Prior Related Work by the Investigators

Published
Last Modified