SERIN: A Scalable Observability System for Actor-Based Distributed Systems

Modern distributed systems increasingly adopt the actor model to achieve scalability and resilience through asynchronous, message-driven concurrency; however, the inherent dynamism and non-determinism of actor interactions pose significant challenges for observability, especially in large-scale deployments. This thesis proposes and evaluates a novel approach to actor-based observability by introducing Serin, a decentralized, meta-actor-based monitoring system specifically designed to address these challenges. Serin decouples the monitoring layer from the primary actor system using Akka Remoting, enabling better fault isolation, modular scalability, and high availability. The system implements log replication through the Raft consensus protocol to ensure consistent, durable telemetry across distributed nodes. Furthermore, it introduces the concept of delegated supervision, a novel policy-driven control mechanism that allows actors to be governed by custom-defined rules without violating their encapsulation. The system architecture is designed to support high-throughput environments through layered monitoring actors, flattened log aggregation models, and lightweight heartbeat-based health checks. These features enable real-time and historical visibility, efficient failure detection, and automated response capabilities through time-based and metric-based policies. By offloading observability responsibilities to an external, isolated actor system, Serin achieves robust performance, contextual monitoring accuracy, and system resilience. Evaluation results under high-volume workloads confirm Serin’s ability to scale, correlate complex event flows, and maintain consistent observability in reactive systems. This work contributes a practical and extensible framework that bridges the gap between theoretical actor concurrency and production-grade observability infrastructure.