Read replicas, often used to manage high read loads in databases, are increasingly seen as outdated due to their inherent complications, such as eventual consistency, increased operational overhead, and infrastructure costs. Dragonfly offers an alternative with its multi-threaded architecture that effectively eliminates the necessity for read replicas by efficiently scaling both vertically and horizontally, thus maintaining low-latency operations without the need for redundant data copies. However, Dragonfly still supports read replicas for specific scenarios where they might be beneficial, such as high availability, connection load balancing, and when vertical scaling reaches its limits. By leveraging Dragonfly’s architecture, organizations can optimize resource usage and potentially enhance system reliability and performance, making read replicas more of a targeted tool rather than a default solution for scaling challenges.