Spanergy brings energy-aware tracing to microservices with low overhead

Spanergy, introduced by researchers César Perdigão Batista, Denis Conan, and Sophie Chabridon, tackles a critical gap in microservices observability: the inability to attribute energy consumption to individual requests. Current distributed tracing tools like OpenTelemetry focus on latency and failures, ignoring energy cost. Spanergy bridges this by correlating per-microservice power measurements with trace spans, creating an energy-aware trace that shows exactly how much energy each request segment (span) consumes. The system works with both synchronous request chains and asynchronous interactions, and follows a rigorous experimental protocol with statistical analysis to validate overhead, conservation, and coverage.

In real-world testing, enabling OpenTelemetry tracing itself increased total experiment energy by 59.1% relative to an uninstrumented baseline. Spanergy's post-processing added only an incremental 15.2% of the baseline energy, meaning the energy-aware tool costs less overhead than the tracing it relies on. Importantly, Spanergy revealed that a non-negligible fraction of request energy comes from spans outside the latency-critical path—insights that can drive energy-efficient microservice design, debugging, and accountability. The approach shows that energy-aware tracing is feasible at modest cost and provides actionable insights for greener cloud applications.