Adaptive Intelligent Tiering for modern storage systems

Lu Pang, Anis Alazzawe, Madhurima Ray, Krishna Kant, Jeremy Swift

Abstract: Highlights•An intelligent tiering mechanism for multi-tiered storage systems is proposed.•Making tiering decisions based on the changing behavior of storage accesses.•Providing data movement candidates using a deep learning model.•A reinforcement learning mechanism is employed to further refine the candidates.•A simulator is developed to simulate the functionality of the tiering system.