Estratégias de escalabilidade para sistemas sob alta demanda: análise comparativa de performance, custo e disponibilidade

Abstract

This work presents a comparative analysis of two widely adopted strategies for increasing the capacity of systems that handle high request volumes: vertical scalability (scale-up) and horizontal scalability (scale-out). The study is based on a systematic literature review that brings together recent findings on performance, elasticity, cost, and availability in distributed environments, and is complemented by an empirical experiment designed specifically for this research. Three scenarios were evaluated using metrics such as throughput, latency, and behavior under increasing load. The results indicate that, although vertical scaling offers relevant performance improvements with lower operational complexity, it is constrained by physical hardware limits and exhibits nonlinear cost growth. Horizontal scaling, on the other hand, demonstrated a greater ability to absorb sudden traffic spikes, achieving significantly higher throughput and more stable latency, while reducing risks associated with single points of failure. However, this approach introduces additional challenges related to distributed consistency, orchestration tools, and operational overhead. The integrated analysis shows that the choice between scaling vertically or horizontally is not universal; instead, it depends on application characteristics, budget constraints, workload predictability, and availability requirements. The study concludes that systems exposed to highly variable or unpredictable demand tend to benefit more from horizontal scaling, whereas smaller or legacy applications may find vertical scaling more adequate. Finally, the work provides practical recommendations for architectural decision-making and suggests future research directions involving hybrid strategies and predictive autoscaling techniques.