Abstract: According to the multirental pattern over virtual network embedding problem in cloud computing， a resource allocation algorithm named MLBVNESDPSO was proposed to reduce the substrate link load， speed up the mapping efficiency and increase the substrate physical resource utilization. The leveraging discrete binary particle swarm optimization algorithm was combined with the virtual network embedding rules in the proposed algorithm. Both CPU and host resources utilization ratio was taken into account， so the physical link bandwidth resource was saved and the time for virtual link mapping process was reduced. The major characteristic of the mapping algorithm was that repeatable mapping for each virtual network could be supported and the load of substrate node could be controlled. Simulation results showed that in the premise of guaranteeing substrate network load the algorithm can achieve better utilization ratio of substrate network and higher revenuecost ratio of virtual networks.

Key words: network virtualization, embedding algorithm, virtual network, integerlinear programming, discrete particle swarm optimization