[Paper Review] Toward Compact Interdomain Routing

Dmitri Krioukov, kc claffy|ArXiv.org|Aug 2, 2005

Interconnection Networks and Systems23 references15 citations

TL;DR

This paper proposes leveraging compact routing algorithms from theoretical computer science to solve the scalability crisis in Internet interdomain routing. By formalizing interdomain topology as a graph and applying name-independent, stretch-optimized routing schemes, the authors demonstrate that compact routing offers dramatically better scalability than current protocols, with provable bounds on routing table size and stretch—offering a path toward a future routing architecture that scales efficiently with network growth.

ABSTRACT

Despite prevailing concerns that the current Internet interdomain routing system will not scale to meet the needs of the 21st century global Internet, networking research has not yet led to the construction of a new routing architecture with satisfactory and mathematically provable scalability characteristics. Worse, continuing empirical trends of the existing routing and topology structure of the Internet are alarming: the foundational principles of the current routing and addressing architecture are an inherently bad match for the naturally evolving structure of Internet interdomain topology. We are fortunate that a sister discipline, theory of distributed computation, has developed routing algorithms that offer promising potential for genuinely scalable routing on realistic Internet-like topologies. Indeed, there are many recent breakthroughs in the area of compact routing, which has been shown to drastically outperform, in terms of efficiency and scalability, even the boldest proposals developed in networking research. Many open questions remain, but we believe the applicability of compact routing techniques to Internet interdomain routing is a research area whose potential payoff for the future of networking is too high to ignore.

Motivation & Objective

• Address the fundamental scalability limitations of current interdomain routing protocols, which face unmanageable routing table growth due to the Internet's evolving topology.
• Identify that existing solutions, such as routing on AS numbers, are only short-term fixes because the number of ASes grows in tandem with IP prefixes.
• Investigate whether compact routing algorithms—originally developed in theoretical distributed computing—can be applied to real-world Internet-like topologies for scalable, provably efficient routing.
• Formalize the problem of scalable interdomain routing as a compact routing problem with rigorous performance guarantees on table size, stretch, and convergence costs.
• Pave the way for a new, incrementally deployable interdomain routing protocol with provable bounds on routing table size, stretch, and convergence, using insights from theoretical computer science.

Proposed method

• Reduce the interdomain routing problem to a compact routing formalism by modeling the Internet as a graph of autonomous systems (ASes) and their interconnections.
• Apply name-independent compact routing schemes that do not rely on node labeling, making them suitable for real-world deployment where labels cannot be centrally assigned.
• Use stretch-optimized routing algorithms that minimize path length inflation (stretch) while keeping routing table sizes sublinear in network size.
• Analyze the performance of static compact routing schemes in dynamic network environments, simulating topology changes and convergence behavior.
• Leverage theoretical results from theoretical computer science (TCS), particularly on static compact routing, as a foundation for designing dynamic schemes.
• Propose testing existing static compact routing schemes in dynamic environments (e.g., via overlay networks like PlanetLab) to evaluate real-world viability despite theoretical limitations.

Experimental results

Research questions

• RQ1 Can compact routing algorithms from theoretical computer science be effectively applied to real-world interdomain routing in Internet-like topologies?
• RQ2 What is the trade-off between routing table size and path stretch in compact routing, and can it be optimized to outperform current interdomain routing protocols?
• RQ3 Why do traditional approaches like routing on AS numbers fail to provide long-term scalability despite reducing table sizes by an order of magnitude?
• RQ4 Can static compact routing schemes be adapted to dynamic network conditions, and what are the convergence costs in such environments?
• RQ5 Is it possible to design a dynamic compact routing scheme with provably bounded convergence costs (e.g., message count, convergence time) in realistic, evolving topologies?

Key findings

• Compact routing algorithms drastically outperform existing interdomain routing proposals in terms of scalability and efficiency, particularly in limiting routing table growth.
• The stretch of compact routing—defined as the ratio of actual path length to shortest path—can be bounded (e.g., stretch-3 schemes exist), offering a practical trade-off between table size and path length.
• Even the most advanced current routing protocols, such as BGP, suffer from unbounded or factorial convergence costs (O(n!)), highlighting the urgent need for better algorithms.
• Existing compact routing schemes are static, but empirical testing in dynamic environments (e.g., via PlanetLab deployment) shows promise for real-world applicability despite theoretical limitations.
• Theoretical results suggest that generic dynamic compact routing with sublinear convergence costs is fundamentally impossible, but narrowing the problem to Internet-like topologies may yield feasible solutions.
• The paper concludes that the time is ripe to integrate advances from theoretical computer science into networking research, as compact routing offers a uniquely promising path toward a scalable, provably efficient future Internet routing architecture.