On modeling knowledge graphs for representing and explaining wide-area distributed storage system

The increasing decentralization of data processing across the computing continuum poses significant challenges for traditional storage infrastructures, which must now operate seamlessly across heterogeneous, geographically distributed environments. Wide-area storage systems address this need by providing a unified data layer that spans multiple physical locations; however, their management becomes increasingly complex as they evolve dynamically in response to workload, infrastructure, and policy changes. This paper addresses the malleability problem in wide-area storage systems, which refers to a system's ability to continuously adapt to changing operational conditions. We propose a knowledge-driven approach based on Knowledge Graphs (KGs) to enable adaptive and intelligent management. The proposed approach models both data and infrastructure layers through a semantic ontology, supports malleability analysis using graph-based metrics, and enables self-adaptive workflows for system reconfiguration. The approach is validated through its integration into DynoStore, a wide-area storage system that manages workloads across multiple locations. Experimental results demonstrate that the KG-based workflow effectively identifies data popularity and infrastructure imbalance, guiding reconfiguration decisions that improve load distribution and resource utilization.