Arquitetura de monitoramento para ambientes de computação em nuvem baseada em consistent hashing

Abstract

Cloud computing is a rapidly growing business model, with significant adoption by individuals and companies seeking a ready-made solution with low initial costs for IT infrastructure components. In this scenario, cloud providers need efficient monitoring systems to ensure high availability, scalability, and compliance with SLAs (Service Level Agreements). However, the dynamic nature of cloud environments, especially in the IaaS (Infrastructure as a Service) model, presents significant challenges due to the high volume of provisioned instances and the constant flow of resource creation and removal. This work presents a distributed monitoring architecture based on the technique of consistent hashing to address these challenges, called Prometheus Ring. This is an operator for Docker Swarm, capable of dynamically distributing the monitoring load among multiple instances of Prometheus, ensuring efficient balancing and resilience to failures. Additionally, the architecture uses Grafana Mimir for long-term storage, allowing the consolidation of historical data without loss of collected metrics. To validate the implementation, a system capable of generating synthetic metrics called Synthetic Exporter was created, which simulates a high-load environment with multiple instances of cloud products. The obtained results demonstrate that the proposed approach reduces monitoring bottlenecks, improves system scalability, and minimizes downtimes, making it a viable alternative for cloud providers needing an efficient solution to monitor cloud computing infrastructures.