Hybrid Cumulative Backup Policies in Reliability Theory  
Author Hongshuang Feng

 

Co-Author(s) Xufeng Zhao; Yilei Bu; Wendi Pang

 

Abstract This paper addresses the operational dilemma faced by 24/7 transactional databases characterized by stochastic update-intensive workloads, where traditional pause-and-backup approaches often result in significant resource underutilization during peak periods. We propose two adaptive hybrid cumulative backup policies: 1) A time-triggered policy executing full backups at periodic intervals KT while performing cumulative backups immediately after each data update; 2) An event-driven policy initiating full backups after N data updates with intermediate cumulative backups. Through a stochastic cost-benefit framework incorporating reliability theory, we derive explicit formulations for expected cost rates that account for system downtime penalties, storage overhead, and recovery complexity. The optimization landscape is systematically explored, yielding closed-form solutions for optimal trigger thresholds K*, K*F , and N*F  that achieve Pareto efficiency between backup frequency and operational disruption. In addition, analytical derivations are rigorously validated through numerical simulations modeling.

 

Keywords Cumulative backup, full backup, database failure, reliability theory
   
    Article #:  RQD2025-344
 

Proceedings of 30th ISSAT International Conference on Reliability & Quality in Design
August 6-8, 2025