Reliability Analysis of SAN Topologies under Variable Workloads Using Hybrid Simulation Techniques

J. Nazdrowicz (Lodz Univ. of Techn., Poland), M. Tuszyńska (Cracow Univ. of Techn., Poland)

Storage Area Networks (SANs) form the backbone of modern data‑center storage infrastructures, where reliability and performance must be maintained under highly variable I/O workloads. Traditional analytical models often provide only coarse estimations of availability, while pure performance simulations fail to capture the stochastic nature of component failures. This paper presents a hybrid simulation framework that integrates Monte‑Carlo reliability modeling with discrete‑time performance simulation to evaluate the behavior of SAN topologies under dynamic, burst‑driven workloads. The proposed method models failures and repairs of switches, links, hosts, and storage controllers, incorporates failover and failback mechanisms, and accounts for capacity degradation during fault conditions. Variable workload patterns, including diurnal variations and burst episodes, are generated using a stochastic traffic model reflecting realistic storage I/O characteristics. Two common SAN architectures—single‑fabric and dual‑fabric multipath configurations—are analyzed in terms of availability, queue dynamics, throughput stability, and delay distributions. Results demonstrate that the hybrid approach exposes complex interactions between failures, traffic surges, and multipath routing, which are not captured by classical analytical availability formulas. In particular, dual‑fabric topologies significantly mitigate performance degradation during failure episodes, reduce tail latency, and exhibit higher operational stability under workload variability. The proposed framework provides a practical tool for evaluating SAN designs, supporting capacity planning, and quantifying the benefits of redundancy in enterprise‑class storage networks.