Wish List for a Topology Generator

An ideal topology generator should enable the use and development of generation models that produce accurate representations of Internet topologies. Thus, it should include features that appeal to the researcher who is in need of accurate synthetic topologies for studying the correctness and performance of protocols and algorithms, as well as to the researcher who is in search for better and more powerful generation models. The following is a list of desirable characteristics for a topology generator.

1. Representativeness. Produces accurate synthetic topologies. Accuracy should be reflected in as many aspects of the actual Internet topology as possible (e.g. hierarchical structure, degree distribution characteristics, etc.).

2. Inclusiveness. Combines the strengths of as many generation models as possible in a single generation tool.

3. Flexibility. Generates topologies over a wide range of sizes. Restrictions such as minimum and maximum number of nodes should be reasonably avoided.

4. Efficiency. Generates large topologies (e.g. number of nodes $> 100,000$) with reasonable CPU and memory consumption.

5. Extensibility. Provides mechanisms that allow the user to easily extend its capabilities by adding new generation models.

6. User-friendliness. Follows the usage principles of standard user interfaces. The user should learn the mechanics of the generation tool only once. For each generation model incorporated in the tool, she should only need to learn the functionality associated with the new model.

7. Interoperability. Provides interfaces to main simulation and visualization applications. It should be possible to generate topologies that can be processed by widely used simulators such as ns [18] and SSF [22].

8. Robustness. Does not sacrifice robustness in the name of efficiency and includes extensive error detection capabilities.

In Section 2.1 we describe the main topology generators and generation models available, in Section 2.2 we discuss some challenges that must be overcome to develop a universal generation tool satisfying our wish list, and in Section 2.3 we argue about a possible approach to tackling those challenges.