Métricas para avaliação de desempenho de um banco de dados distribuído

Abstract

The choice of data storage technology is crucial to ensure not only performance but also the availability and integrity of information, especially in high-demand scenarios. This study compared PostgreSQL in its conventional and distributed versions using HammerDB as a benchmark, presenting quantitative results that highlight the significant performance differences between these approaches. Under a load of 150 users, the manually tuned PostgreSQL configuration achieved approximately 30,000 transactions per minute, showing a notable improvement over the standard PostgreSQL installation. However, the distributed version of PostgreSQL significantly outperformed these benchmarks, reaching an average of 55,000 transactions per minute. These results underscore the effectiveness of the distributed approach in high-demand environments, demonstrating its superiority in terms of processing capacity and response to requests. The metrics defined in this work provide clear guidelines for the implementation and optimization of PostgreSQL, allowing for an informed choice of the most suitable technology for different data storage scenarios. Thus, the findings not only confirm the superiority of the configured version over the standard installation but also reinforce the advantage of distributed PostgreSQL for applications requiring high performance and continuous availability.