Network topology optimization using Tabu Search

Designing a network with a certain level of survivability can be modeled as an instance of the Generalized Steiner Problem with vertex connectivity constraints (GSP). In this work we use the Tabu Search (TS) metaheuristic to find low cost topologies that satisfy the vertex connectivity constraints. The basic building blocks that TS needs are: a feasible initial solution, and a move and neighborhood definition. The move is specified as an exchange of paths, therefore the neighborhood is path-based. The initial solution is constructed using a greedy algorithm. The software implementation time was reduced using reusable open source libraries and frameworks such as OpenTS, JUNG and Jakarta Commons Collections. Furthermore, the JUnit framework was used for unit testing to support every step of the implementation.