An NFV-based Scheduling and Flexible Deployment Scheme for RaaS Functions in Cloud Data Centers

Recommender systems appear on many commercial websites because E-Commerce and social websites are popular. However, many businesses do not have enough capacity to develop their recommender systems, which must be outsourced by cloud-based service providers because the amount and complexity of data increase. Recently, network function virtualization (NFV) has been proposed to use virtualization technology to replace dedicated network equipment for software network functions. Virtual network functions (VNFs) run on commodity servers or standard physical machines. In this paper, an NFV-based scheduling and flexible deployment scheme (SFDS) is proposed to consider the characteristics of recommendation-as-a-service (RaaS), including the popularity and execution order of the functions. According to the popularity of the functions, a dynamic deployment algorithm is able to flexibly create or remove the VNFs on the virtual nodes. Finally, the simulation results show a shorter completion time. SFDS enhances the resource utilization of the system and has a better successful reception rate.