Cost Efficient State-aware Function Placement and Flow Scheduling for NFV Networks

Via the newly emerging technologies Network Function Virtualization (NFV), network service providers can acquire cloud resources to provision network services (e.g., network service chain) in a flexible and on-demand manner, thus significantly reducing both the capital and operational expense. In this case, how to deploy Virtual Network Functions (VNFs) has become a critical issue and has attracted much attention in the literature. Existing VNF placement studies usually ignore the "state" of VNFs and assume that all VNFs are stateless. However, it is well known that some VNFs are stateful and a collection of flows may share an aggregate state. How to deploy such stateful VNFs is still under-investigated. In this paper, we are motivated to investigate how to add a stateful VNF into an existing service chain in a cost-efficient manner. In particular, we intend to maintain the dynamic aggregate state on the exclusive server to avoid state synchronization by assigning the aggregate flows to the same VNF node. We first formulate the problem into an integer linear programming (ILP) form, which is then proved to be NP-hard. To address the computation complexity, we then propose a dynamical programming based heuristic algorithm, whose high efficiency is extensively proved via trace-based simulations. The performance evaluation results show that our heuristic algorithm performs close to the optimal solution and outperforms a greedy-based competitor. In addition, we also show that the state of VNF has deep impact on the VNF deployment cost.