An Algorithm for Unsplittable Flow Problem in Flexible Reconfigurable Network

Information infrastructures such as Internet have been a great success in the past few decades and gradually become mature, diversified user requirements continue to grow at a fabulous speed. Multiplex services on secure networked environments are now deemed as an application licensed to customers for use as a trustworthy service on demand. The flexible reconfiguration network is the next step to provide pervasive trustworthy services. A major challenge is the efficient assignment of substrate resources for unsplittable flows which is called unsplittable flow problem (UFP). The problem is an NP-hard variant of max flow and a generalization of single-source edge-disjoint paths with applications to scheduling, load balancing and virtual-circuit routing problems. Current constant-factor approximation algorithms for several natural optimization versions of the problem have been provided but are not suitable for providing multiple services due to the shortest path. In this paper, we first model the problem as an undirected graph with multiple appropriate subroutines, and then provide a load-balance based resource assignment algorithm for UFP in flexible reconfiguration network. The algorithm tends to make efficient use of the substrate resource by choosing the route which has the minimal average utilization based on load balance, which will leave more resources for future requests and in turn improve the resource utilization. Our simulation experiments demonstrate that the algorithm can achieve good performance.