Solving the Virtual Network Mapping Problem with Construction Heuristics, Local Search and Variable Neighborhood Descent

The Virtual Network Mapping Problem arises in the context of Future Internet research. Multiple virtual networks with different characteristics are defined to suit specific applications. These virtual networks, with all of the resources they require, need to be realized in one physical network in a most cost effective way. Two properties make this problem challenging: Already finding any valid mapping of all virtual networks into the physical network without exceeding the resource capacities is NP-hard, and the problem consists of two strongly dependent stages as the implementation of a virtual network's connections can only be decided once the locations of the virtual nodes in the physical network are fixed. In this work we introduce various construction heuristics, Local Search and Variable Neighborhood Descent approaches and perform an extensive computational study to evaluate the strengths and weaknesses of each proposed solution method.