Efficient Auto-Scaling Approach in the Telco Cloud Using Self-Learning Algorithm

Abstract

Network Function Virtualization (NFV) and Software Defined Network (SDN) technologies makes it possible for the Telco Operators to assign resource for virtual network functions (VNF) on demand. Provision and orchestration of physical and virtual resource is crucial for both Quality of Service (QoS) guarantee and cost management in cloud computing environment. Auto-scaling mechanism is essential in the lifecycle management of those VNFs. Threshold based policy is always applied in classic IT cloud environments which can not satisfy carrier grade requirements such as reliability and stability. In this paper, we present a novel SLA-aware and Resource-efficient Self-learning Approach (SRSA) for auto-scaling policy decision. The scenarios of the service volatility is categorized into daily busy-and-idle scenario and burst-traffic scenario. First, we formulate the workload of the VNF as discrete-time series and treat procedure of policy-making in auto-scaling as a Markov Decision Process (MDP). Second, parameters in the Reinforcement Learning process are tuned cautiously. Finally the experiments show that our solution outperforms threshold based policy and voting policy adopted by RightScale in oscillation suppression, QoS guarantee, and energy saving.