Nesterov Accelerated Gradient-based Algorithm for the Allocation Problem of NFV Service

Zhonghui Wu, Xingyan Chen, Changqiao Xu, Shujie Yang

Abstract: Network Functions Virtualization, as well as Software Defined Networking, is a crucial technology to distributed cloud network, providing support for network services allocations with elastic Virtual Network Functions instantiated in a flexible way at distributed cloud locations. In this paper, an O (E) approximation optimal solution with time complexity O([1/(E)]) to NFV Service Allocation Problem is proposed. Enlightened by the Nesterov Accelerated Gradient-based method, we use the cloud augmented graph model and redefine the service chain model to deal with the flow requirement, processes and transmission of cloud network. We formulate the VNFs services place problem as a non-smooth concave problem with integer programming form. Then, we translate the integer form into the linear program by linear relaxation of the problem. We further propose a Distributed Nesterov Accelerated Gradient-based algorithm which enables network nodes to obtain the optimal scheme by only requiring weight information from one-hop neighbors. Besides, we give three main theoretical results of our algorithm, which are queue-length reduction performance, utility optimal and convergence rate. Finally, multiple simulations are conducted to show the dominance of the proposed algorithm against the state-of-art solution.