Network Function Placement for Service Chains with Server Maintenance Cost

Abstract

Network function virtualization (NFV) is a technique of building new network architecture that can decouple network functions from proprietary equipment and thus increase the flexibility of network services deployment while reducing the total cost significantly. In the NFV scheme, network functions are placed on generic servers over the network. When a network service is requested, it will be implemented by chaining several Virtual Network Functions (VNFs) placed on servers to form a Service Function Chain (SFC). The deployment of SFC faces an unprecedented challenge: when several network services are requested, which servers should be chosen to form service function chains while minimizing the total cost? Many existing optimization models consider the feasibility of placing the SFCs or minimizing the resource consumed by the VNFs. However, activate and maintain a server to hold VNFs also consume a huge amount of energy, and it's possible to lower the total energy cost by carefully arrange the servers with low maintenance cost. Therefore, in this paper, to investigate the network function placement for service chains with server maintenance cost, we propose a heuristic algorithm, which firstly construct a server set that may accommodate all service chains with low server maintenance cost. Then it formulates a linear programming to find the schemes of chaining the VNFs (also called placement configurations) with minimum communication cost in the server set, and obtain its optimal fractional solutions. Eventually, we perform a rounding scheme to achieve integer solutions that out-performs the greedy scheme in many evaluations.