Investigação dos efeitos do envelhecimento de software em sistemas de gerenciamento de banco de dados

Abstract

Database management systems are designed to ensure high availability and reliability. In these systems, software aging, a phenomenon in which long-running systems experience gradual performance degradation or progressive resource consumption, remains underexplored, especially in NoSQL systems. This study systematically investigated, in long-term experiments, aging effects in six major open-source relational and non-relational DBMSs. We introduced the SWCycles approach, to accelerate aging through repeated stress-wait cycles under extreme workload. Experiments consisted of continuous read operations with 500 concurrent virtual users during 48-hour stress periods followed by 6-hour waits, over 131.25 and 171.25 days. System resources were continuously monitored, and time series were analyzed using the non-parametric Mann-Kendall test. Results showed that the most evident symptom occurred in RAM, with slow cumulative growth of consumption over time. The finding confirm that software aging does not manifest abruptly, but normally as a gradual trend in resource usage, which can lead to performance degradation and service unavailability.