Optimization Method for Cross-Domain Heterogeneous Storage System Management Based on Improved NSGA-III

Given the diversity and geographic distribution of data resources across multiple data centers, achieving unified storage and resource management presents significant challenges, complicating efforts to meet the demands for efficient cross-domain distributed data access. To address these challenges, this study proposes a cross-domain heterogeneous storage cluster architecture. Leveraging Alluxio as the foundational storage layer for data-intensive applications, this architecture facilitates unified management across various storage systems. For the storage system selection problem, this paper adopts a multi-objective approach with minimum cost, maximum reliability, shortest access distance, and load balancing as the target functions. An improved NSGA-III algorithm, incorporating generalized inverse learning and adaptive evolutionary mechanisms, is proposed for intelligently selecting the optimal storage scheme. The reasonableness and effectiveness of the improved algorithm are validated through a comparative analysis of the convergence and diversity metrics of the original and improved NSGA-III algorithms on benchmark problems, along with assessments of each objective's adaptability in real-world application scenarios.