In this paper we focus on the
unsplittable flow problem
(UFP): given a directed network with arc capacities and a set of connections (requests) defined by origin node, destination node and bandwidth requirement, find a subset of the connections of maximum total demand for which each connection uses only one path and the sum of demands crossing the arc does not exceed its capacity. The UFP can be applied in survivable connection-oriented network (e.g. MPLS) for assignment of backup paths of failed connections. Since the UFP is NP-complete, we propose two new effective heuristic algorithms for the UFP. We evaluate the performance of proposed schemes by making a comparison with their counterparts using various network topologies and demand patterns. Obtained results indicate that proposed heuristics provide substantial improvement comparing to existing methods.