O-RAN Edge Telemetry: Multi-Destination Data Pipeline

Streamline O-RAN network observability with a single edge pipeline that collects telemetry from DU/RU/CU nodes and routes data simultaneously to multiple destinations: real-time dashboards, analytics platforms, and long-term storage.

The Problem

Telco operators managing O-RAN networks face a complex observability challenge:

• Thousands of distributed edge nodes generate continuous telemetry (DU, RU, CU units)
• Multiple consumption patterns require the same data:
  • Real-time dashboards (Grafana) for NOC operations
  • Long-term storage (Parquet) for capacity planning
  • Analytics platforms (Cloudera) for ML/AI insights
  • Compliance reporting for PTP timing requirements
• The edge gap: Data needs to be collected, shaped, and enriched before it reaches your Cloudera/Kafka/Grafana backends — but there's no lightweight, Red Hat-integrated way to do it at the edge

The result: 3-5x network overhead from shipping raw data, inconsistent schemas across destinations, and no edge-native processing before data hits your backends.

The Solution

Single collection, multiple destinations: Expanso Edge pipelines run on OpenShift Single Node OpenShift (SNO) directly alongside RAN workloads, collecting O-RAN telemetry once and routing to all destinations simultaneously.

DU Telemetry (PTP, CPU, PRB, RSRP)

↓

Expanso Edge Pipeline

parse, normalize, enrich, filter anomalies, fan-out

↓

📊 Grafana / Prometheus

↓

📦 Parquet Storage

↓

☁️ Cloudera CDP

↓

💾 Local Buffer

Where Expanso Fits In Your Stack

Your backend stays exactly as-is — Cloudera CDP, Kafka, Grafana, HDFS/Ozone. Expanso Edge handles what happens before data reaches those systems: collect, shape, augment, and deliver to multiple destinations simultaneously.

Layer	What	How Expanso Helps
Collect	MQTT, OPC-UA, syslog, file tailing from DU/RU/CU nodes	Single lightweight agent on OpenShift SNO — no NiFi deployment at the edge
Shape	Parse messy telco logs, normalize schemas, filter noise	Bloblang processors run at the edge — only clean data leaves the site
Augment	Enrich with cell site metadata, geo coordinates, compliance zones	Lookups and enrichment happen before transmission, not after
Deliver	Fan-out to Kafka, Grafana, Parquet, Cloudera — all at once	One pipeline, multiple destinations. Your backends receive ready-to-use data

Tower → Expanso Edge (collect, shape, augment) → Kafka / Cloudera / Grafana / S3
        ├── runs on OpenShift SNO                  (your existing backends, unchanged)
        ├── buffers locally during outages
        └── managed via Red Hat-integrated operator

Key Benefits

• Edge-native processing: Collect and transform at the source, not in the datacenter
• 99% bandwidth reduction: Only shaped, filtered data leaves the edge site
• Zero data loss: Local buffering handles burst traffic and connectivity drops
• Unified data model: Every destination receives consistent, enriched schemas
• Multi-destination fan-out: Same data to Kafka, Grafana, Parquet, and Cloudera simultaneously
• Red Hat integrated: Runs as certified OpenShift operator on existing SNO infrastructure

What You'll Build

This guide walks through creating a production-ready O-RAN telemetry pipeline that:

1. Collects PTP timing, CPU, PRB utilization, and RF metrics from DU nodes
2. Transforms raw telemetry with Bloblang processing (compliance classification, enrichment)
3. Routes to multiple destinations using fan-out broker pattern
4. Monitors pipeline health with built-in observability

Key Metrics Processed

Metric	Source	Purpose	Compliance Threshold
PTP4L Offset	DU Timing	5G sync compliance	less than ±100ns (compliant), ±1000ns (critical)
PRB DL/UL %	DU Scheduler	Resource utilization	greater than 90% (congested)
CPU %	DU System	Performance monitoring	greater than 80% (alert)
RSRP/SINR	UE Reports	RF quality	RSRP less than -120dBm (poor coverage)

Prerequisites

• Expanso Edge running on OpenShift SNO nodes
• Access to DU telemetry endpoints or files
• Grafana + OTEL Collector + Prometheus stack
• Parquet writer capability
• Cloudera Data Platform (CDP) or Kafka endpoint

Get Started

Choose your path:

Interactive Explorer

See each O-RAN telemetry processing technique with side-by-side transformations

Step-by-Step Tutorial

Build the pipeline incrementally:

1. Collect O-RAN Metrics
2. Transform and Enrich
3. Multi-Destination Routing
4. Grafana Dashboards
5. Parquet and Cloudera

Complete Pipeline

Download the production-ready solution