| Summary: | The Internet at the interdomain level is a complex network of approximately 50,000 Autonomous Systems (ASes). ASes interconnect through two types of links: (a) transit (customer-provider) and (b) peering links. Recent studies have shown that despite being optional for most ASes, a rich and dynamic peering fabric exists among ASes. Peering has also grown as one of the main instruments for catching up with asymmetric traffic due to CDNs, online video traffic, performance requirements, etc. Moreover, peering has been in the spotlight recently because of peering conflicts between major ISPs and Content Providers. Such conflicts have led to calls for intervention by communication regulators and legislation at the highest levels of government. Peering disputes have also sometimes resulted in partitioning of the Internet.
Despite the broad interest and intense debate about peering, several fundamental questions remain elusive. The objective of this thesis is to study peering from a techno-economics perspective. We explore the following questions:
1- What are the main sources of complexity in Internet peering that defy the development of an automated approach to assess peering relationships?
2- What is the current state of the peering ecosystem, e.g., which categories of ASes are more inclined towards peering? What are the most popular peering strategies among ASes in the Internet?
3- What can we say about the economics of contemporary peering practices, e.g., what is the impact of using different peering traffic ratios as a strategy to choose peers? Is the general notion that peering saves network costs, always valid?
4- Can we propose novel methods for peering that result in more stable and fair peering interconnections?
We have used game-theoretic modeling, large-scale computational agent-based modeling, and analysis of publicly available peering data to answer the above questions. The main contributions of this thesis include:
1- Identification of fundamental complexities underlying the evaluation of peers and formation of stable peering links in the interdomain network.
2- An empirical study of the state of the peering ecosystem from August 2010 to August 2013.
3- Development of a large-scale agent-based computational model to study the formation and evolution of Internet peering interconnections.
4- A plausible explanation for the gravitation of Internet transit providers towards Open peering and a prediction of its future consequences.
5- We propose a variant of the Open peering policy and a new policy based on cost-benefit analysis to replace the contemporary simplistic policies.
|
|---|
