An efficient binary-search based heuristic for extended Unsplittable Flow Problem

Routing a given set of flows with bandwidth requirements is a fundamental problem, which has been formulated as Unsplittable Flow Problem (UFP). One of the key issues in this formulation is that the bandwidth requirement is fixed for each flow. In practice, however, many applications (e.g., video-on-demand, backup and replication services) would greatly benefit from getting more bandwidths beyond the fixed minimum bandwidth requirement. To achieve this, we extended UFP such that bandwidth requirements will be given in a range. The challenge is now how to find paths and dynamically determine the allocated bandwidths such that we can provide more bandwidth to each flow while making sure that each flow gets at least the requested minimum amount. Like UFP, the extended UFP problem is also NP-Hard. Therefore, we propose two heuristics and demonstrate their efficiency using simulation.