Robust Network Planning

Building a network encompasses many tasks: from network planning to hardware installation and configuration, to ongoing maintenance. In this chapter we focus on the process of network planning. It is possible (though not always wise) to design a small network by eye, but automated techniques are needed for the design of large networks. The complexity of such networks means that any “ad hoc” design will suffer from unacceptable performance, reliability and/or cost penalties. Network planning involves a series of quantitative tasks: measuring the current network traffic and the network itself; predicting future network demands; determining the optimal allocation of resources to meet a set of goals; and validating the implementation. A simple example is capacity planning: deciding the future capacities of links in order to carry forecast traffic loads, while minimizing the network cost. Other examples include traffic engineering (balancing loads across our existing network) and choosing the locations of Points-of-Presence (PoPs) though we do not consider this latter problem in detail in this chapter because of its dependence on economic and demographic concerns rather than those of networking. Many academic papers about these topics focus on individual components of network planning: for instance, how to make appropriate measurements, or on particular optimization algorithms. In contrast, in this chapter we will take a system view. We will present each part as a component of a larger system of network planning. In the process of describing how the various components of network planning interrelate, we observe several recurring themes: