Routing management in very large-scale networks

Reviewer: Thomas Narten

This paper focuses on two main ideas. First, the author presents a summary of related works in hierarchical routing, but criticizes them because they use routing algorithms that are not loop-free. A hierarchical architecture is then defined. Second, the paper applies a new routing algorithm (described by the author in a separate paper) to this architecture and compares the complexity of the routing management for this new algorithm to that of other common routing algorithms. The complexity analysis shows that the new algorithm is not worse than (and in several cases is better than) other routing architectures. This result is not unexpected, however, as the author had previously shown that the new algorithm is superior to the other algorithms in the context of a flat namespace. The paper gives a good summary of related hierarchical routing architectures, but its presentations of the specific architecture and of how the new routing algorithm uses it are incomplete and difficult to follow. The practicality of the architectures is difficult to evaluate, as the author leaves it to the reader to discover important details such as how regions are constructed and how routing information flows between them. Moreover, the analysis of the routing algorithm does not rely on exact details of the architecture; it uses general terms common to hierarchical routing architectures instead, raising the question of why the architecture is presented in so much detail. The analysis itself is interesting, however, focusing on communication, time, and storage complexity.