Telemetry: A Packet Centric Lens

Overview

Understanding network behavior at the packet level provides unprecedented visibility into performance bottlenecks and optimization opportunities.

Note

This project develops packet-centric telemetry tools that integrate with EdgeRIC for fine-grained network analysis.

Motivation

Traditional telemetry approaches aggregate data, losing critical timing and ordering information. A packet-centric approach preserves:

  • Per-packet latency distributions

  • Queue dynamics and buffer occupancy

  • Scheduling decision impacts

  • Cross-layer correlations

Key Features

Real-time Packet Tracing

Capture and analyze packets as they traverse the RAN stack with minimal overhead.

Causal Analysis

Link packet-level events to scheduling decisions and network conditions.

Visualization Tools

Interactive dashboards for exploring packet traces and identifying patterns.

Integration with EdgeRIC

The telemetry system provides:

  • Input features for ML-based μApps

  • Ground truth for policy evaluation

  • Debugging and diagnostics capabilities

Results

Coming soon