Thanks to Network Function Virtualization (NFV), Internet Service Providers (ISPs) can improve network resource utilization with significantly reduced capital and operational expenditures. To dig deeper into the potential of NFV, an important challenge is the resource allocation problem in NFV (NFV-RA), which can be divided into three stages: VNFs chain composition, VNF forwarding graph embedding, and VNFs scheduling. The key to the NFV-RA problem is to design an effective and coordinated resource allocation algorithm for the three stages. Besides, the NFV-RA problem has been proved to be NP-Hard, and thus most existing approaches focus on heuristic and meta-heuristic algorithms. In this paper, we propose an NFV online coordinated resource allocation framework (OCRA) that completes the three stages simultaneously in a coordinated manner by combining parallel Multi-Agent Deep Reinforcement Learning with novel neural networks and RL training techniques. The extensive experimental results show that compared with the state-of-the-art solutions, OCRA is highly-efficient in terms of time, with up to 50% and 10.8% improvement on resource overhead and acceptance ratio, respectively.

As a novel evolutionary algorithm, fruit fly optimization algorithm (FOA) has received great attentions and wide applications in recent years. However, existing literature have demonstrated that the basic FOA often risks getting prematurely stuck in the local optima. In this paper, an improved FOA, named as dynamic generalized opposition-based learning fruit fly optimization algorithm (DGOBL-FOA), is proposed to mitigate the aforementioned drawback hence improve the optimization performance. Three carefully designed operators are incorporated into the basic FOA, i.e., a cloud model based osphresis search is applied to enhance the local refinement search ability in the osphresis phase, then a generalized opposition-based learning operation is adopted to strengthen the global coarse search ability, meanwhile a dynamic shrinking parameter strategy is designed to adjust the learning intensity and narrow down the search space iteratively, which contributes to a good balance between the global exploration and local exploitation. To verify the effectiveness of the proposed algorithm, numerical experiments are conducted on 18 well-studied benchmark functions with dimension of 30. The computation results and statistical analysis indicate that the proposed DGOBL-FOA achieve significantly better performance comparing to other FOA variants and the state-of-the-art metaheuristics.