Electrical and Computer Engineering ETDs

Author

Joud Khoury

Publication Date

2-9-2010

Abstract

In less than three decades, the Internet was transformed from a research network available to the academic community into an international communication infrastructure. Despite its tremendous success, there is a growing consensus in the research community that the Internet has architectural limitations that need to be addressed in a effort to design a future Internet. Among the main technical limitations are the lack of mobility support, and the lack of security and trust. The Internet, and particularly TCP/IP, identifies endpoints using a location/routing identifier, the IP address. Coupling the endpoint identifier to the location identifier hinders mobility and poorly identifies the actual endpoint. On the other hand, the lack of security has been attributed to limitations in both the network and the endpoint. Authentication for example is one of the main concerns in the architecture and is hard to implement partly due to lack of identity support. The general problem that this dissertation is concerned with is that of designing a future Internet. Towards this end, we focus on two specific sub-problems. The first problem is the lack of a framework for thinking about architectures and their design implications. It was obvious after surveying the literature that the majority of the architectural work remains idiosyncratic and descriptions of network architectures are mostly idiomatic. This has led to the overloading of architectural terms, and to the emergence of a large body of network architecture proposals with no clear understanding of their cross similarities, compatibility points, their unique properties, and architectural performance and soundness. On the other hand, the second problem concerns the limitations of traditional naming and discovery schemes in terms of service differentiation and economic incentives. One of the recurring themes in the community is the need to separate an entity's identifier from its locator to enhance mobility and security. Separation of identifier and locator is a widely accepted design principle for a future Internet. Separation however requires a process to translate from the identifier to the locator when discovering a network path to some identified entity. We refer to this process as identifier-based discovery, or simply discovery, and we recognize two limitations that are inherent in the design of traditional discovery schemes. The first limitation is the homogeneity of the service where all entities are assumed to have the same discovery performance requirements. The second limitation is the inherent incentive mismatch as it relates to sharing the cost of discovery. This dissertation addresses both subproblems, the architectural framework as well as the naming and discovery limitations.

Keywords

Computer network architectures., Internet--Technological innovations., Data mining., Internet--Security measures., Wireless Internet., Internet--Economic aspects.

Document Type

Dissertation

Language

English

Degree Name

Computer Engineering

Level of Degree

Doctoral

Department Name

Electrical and Computer Engineering

First Committee Member (Chair)

Heileman, Gregory

Second Committee Member

Shu, Wei

Third Committee Member

Krause, Kate

Share

COinS