S-iNAS: A Performance-Centric SRN Modeling and Analysis of Time and Event Based Scaling Strategies in Ceph-Based Industrial Network-Attached Storage

Emerging Industry 4.0 applications and digital twin paradigms generate massive, highly dynamic data streams that pose significant challenges for storage infrastructures in industrial environments. In this study, we address the issue of scalable performance by introducing S-iNAS, a scalable Ceph-based framework capable of adapting to sharply fluctuating workloads while maintaining high throughput and low latency for industrial network-attached storages (iNAS). Our approach leverages a SRN model to rigorously capture key performance factors: concurrent read/write workflows, replication overhead, and dynamic scaling triggers for Ceph Object Storage Daemons (OSDs). We investigate two principal Scaling Policies—time-based (scheduled expansions) and event-based (threshold-triggered)—under varied arrival rates, client concurrency levels, VM instantiation delays, and read/write compositions. Experimental results from extensive Sensitivity Analyses demonstrate that time-based scaling, although maintaining throughput stability under moderate loads, risks delayed responsiveness when demands spike. By contrast, event-based scaling offers agile adaptation and mitigates latency at the cost of more frequent reconfigurations. These findings highlight how even small modifications to scale-out thresholds or scheduling intervals can substantially influence performance. Compared with existing Petri Net or autoscaling research, our SRN driven modelling helps practitioners comprehend Ceph’s replication and I/O concurrency behaviors, thereby filling a critical gap in performance-centric solutions for iNAS. Ultimately, S-iNAS equips practitioners with actionable guidelines to fine-tune scaling policies, ensuring robust, cost-effective, and high-throughput data services for mission-critical industrial workflows and digital twin ecosystems.