Objective

Implement selective profiling for real-time monitoring, minimizing overhead, and pushing metrics to a Grafana dashboard to optimize performance and support debugging across scalable sets of nodes.

Key Goals

• Selective Profiling: Efficient instrumentation of specific code sections, avoiding performance hits for non-profiled parts. Similar to dynamic selection of log-events.
• Low Overhead: Profile only necessary sections at compile-time using OCaml PPX, with alternatives like dynamic linking under evaluation.
• Scalable Event Monitoring: Push profiling events via OpenTelemetry to visualize and compare traces in Grafana.

Key Deliverables and Milestones

• Octez performance can be observed with low-overhead selective profiling
• Observation can be done using Grafana dashboards %Performance Monitoring Visualisation.
• Profiling tools are deployed on monitoring infrastructure
• Profiling is extended to bandwidth and P2P