Brief Announcement: Understanding Self-stabilizing Node-Capacitated Overlay Networks Through Simulation

Abstract

Overlay networks, where connections are made over logical links composed of zero or more physical links, are a popular paradigm in modern distributed computing. The use of logical links allows the creation of a variety of network topologies with desirable properties such as low degree and low diameter, regardless of the (usually) fixed physical topology. Many of these overlay networks operate in unfriendly environments where transient faults are commonplace. Self-stabilizing overlay networks present one way to manage these faults. In particular, self-stabilizing overlay networks can guarantee that the desired network topology is created when starting from any weakly-connected initial state.

To date, work on self-stabilizing overlay networks has assumed the network has either unbounded bandwidth, or that the bandwidth constraints are placed on the communication links themselves. In practice, however, the bandwidth constraints are actually capacities on the nodes: adding and deleting logical links does not change the fixed physical links being used. In this work, we describe the node-capacitated model for self-stabilizing overlay networks. To better understand this new model, we created a simulation and ran it numerous times while adjusting various parameters. We discuss this simulation and several experiments. Finally, we propose future directions for self-stabilizing node-capacitated overlay networks.

This project is funded in part by the Louis Stokes Alliance for Minority Participation (LSAMP).