From Self-Stabilization to Self-Optimization: Principles of Distributed Network Design

Abstract

The tutorial provides an overview of the principles and "pearls'' of constructing, maintaining and optimizing network topologies, such as peer-to-peer overlays or datacenter interconnects. The tutorial consists of two parts: Self-stabilization: In the first part, we start simple by discussing basic mechanisms to reconfigure networks while maintaining connectivity. You will then learn how to design and analyze self-stabilizing algorithms for line topologies as well as more sophisticated networks such as hypercubes. A self-stabilizing algorithm guarantees that the network will automatically reconfigure to a "good topology" from an arbitrary initial state. Self-optimization: After we have learned how to design self-stabilizing networks, we will consider algorithms to design self-adjusting resp. "self-optimizing" networks: networks which not only "repair" themselves but also optimize themselves towards the demand. The study of such self-adjusting networks is motivated by emerging technologies in the context of datacenters and wide-area networks, allowing to reconfigure networks at runtime.