Smart Energy-Aware Data Allocation for Heterogeneous Memory

Heterogeneous memory is an emerging concept that uses multiple kinds of memories to balance the contradictions between the performance and cost. Many prior researches have addressed the heterogeneous memory designs, which attracted both researchers and practitioners. However, the main challenge in memory design is that the moving costs between memories cannot be ignored, such that forming an effective data allocation plan is a critical part for reducing total costs. This paper focuses on this issue and proposes an adaptive energy-aware approach that is designed for efficiently generating adaptive data allocation plans in heterogeneous memories. Two crucial variables are considered the condition constraints, including time and energy consumptions. The proposed method is called Smart Data Allocation Model (SDAM), which is supported by our algorithm, Smart Switch Allocation Algorithm (SA2). The experimental results have depicted that our approach is superior to prior developed methods. The likelihood of generating optimal solutions is 80.4% - 90.1%.