ML-Based Adaptive Backup Snapshot Scheduling For Storage Area Networks

Authors

  • Sarim Yaseen Khan Ilma University Karachi
  • Raja Sohail Ahmed Larik School of Computer Science and Engineering, Nanjing university of science and technology PR China
  • Allah Bachayo Government College University Hyderabad, Pakistan
  • Basit Ali Ziauddin University Faculty of Engineering, Science, Technology& Management
  • Kanwal Ilma University Karachi

DOI:

https://doi.org/10.54692/lgurjcsit.v10i1.203

Keywords:

Machine Learning, Adaptive Systems, Random Forest, Gradient Boosting, Recovery Point Objective

Abstract

In this research study, the Adaptive Predictive Snapshot Scheduler (APSS) is introduced as a new intelligent solution that uses machine learning to optimize backup snapshot scheduling in SAN environments dynamically. One of the most basic parts of enterprise data infrastructure is the Storage Area Network (SAN), and backup snapshots are an operationally difficult SAN feature that has not been widely discussed. Traditional scheduling solutions rely on fixed rules and fail to account for the dynamic, unpredictable nature of enterprise workloads. This rigidity results in suboptimal recovery windows, high storage costs, increased risk of data loss, and network performance issues during peak traffic periods. The APSS framework performs classification with Random Forest and regression with Gradient Boosting in real time, leveraging a hybrid ML approach to achieve high accuracy. The system then dynamically moves towards an optimal configuration of snapshot frequency, timing, and resource allocation based on these predictions, which reduces disruption and unnecessary data captures and ensures the compliance of the Recovery Point Objective. The proposed methodology is tested in a simulated SAN environment with enterprise workload parameters for 90 days. A comparison is made with three different base scheduling strategies: fixed-interval, threshold-based, and time-of-day scheduling. Results show that the APSS framework reduces the average backup completion time by 34.5%, storage overhead by 28.3%, and peak bandwidth consumption by 22.6%, and increases the Recovery Point Objective (RPO) compliance rate by 41.2%, compared to the best-performing baseline method.

Downloads

Published

08-06-2026

Issue

Section

Articles