NEST: Optimal deploying DAG-SFCs to maximize the flows wholly served in the network edge

Abstract

Compared with the dedicated hardware-based middle-boxes, Network Function Virtualization (NFV) makes network services more flexible, scalable, and cost-efficient by replacing the network functions, e.g., Firewalls, Deep Packet Inspections (DPI), Load Balancers, by Virtual Network Functions (VNFs) running on general-purpose commercial servers. Multiple VNFs usually chain up to formulate Service Function Chains (SFC) to enable complex network services. With the exploitation of VNF parallelism, SFC with parallel VNFs (Para-SFC) is proposed to reduce the SFC execution delay. However, the inner-structure of a Para-SFC is usually complex and a Para-SFC may have multiple legitimate topologies. DAG-SFC is one of the basic sharping methods for standardizing the topology of Para-SFCs. To adapt to the fast development of edge computing, it is critical to find an optimal deployment of Para-SFCs in hybrid edge-and-cloud network with the purpose of maximizing the amount of the flows completely serviced in the edge. In this paper, we first analyze the target problem via integer program modeling and prove its NP-hardness. Then, we propose a heuristic algorithm named NEST, which is based on the combination of maximum spanning tree and Next Fit. With extensive evaluations, we demonstrate that, compared with the benchmark methods, NEST can achieve up to respective 14% and 37% performance gain on DAG-SFC edge deployment maximization while can always realize (up to respective 276% and 305%) higher efficiency of link usage.