Develop Telco-Grade Network Observability Platform (5G SA, Wireline, eBPF, Kafka & ClickHouse)
Budget: $30 – $250 CAD
I am seeking to build a production-grade, zero-blind-spot Network Observability Platform. The platform will ingest, decode, and visualize massive telemetry streams from both mobile (4G/5G SA) and wireline telecom infrastructure.
Instead of heavy legacy sidecars, the solution must use eBPF (extended Berkeley Packet Filter) for host/container instrumentation, routing high-capacity XDR streams into an AI-ready data lake.
Detailed Scope of Work & Requirements
1. Infrastructure & Probe Deployment (Cloud-Native)
Kubernetes/NFV Instrumentation: Create IaC (Helm/Terraform) to deploy eBPF collection agents across multi-tenant Kubernetes, Red Hat OpenShift, and legacy NFV nodes.
Kernel Integration: Configure kprobes and uprobes to trace traffic safely inside the host Linux kernel, targeting 5G Network Slices and core workloads without exceeding 1.5% host CPU overhead.
High Availability: Design the agents for a 24/7 mission-critical operational environment with automated self-healing loops.
2. Protocol Decoding Engine & High-Capacity Pipeline
Protocol Decoding: Build custom parsing layers within the data pipeline to decode Mobile (GTP-U, PFCP, HTTP/2, SBA) and Wireline (PPPoE, DHCP, QoS, BNG/BRAS) protocols.
Kafka Streaming Tier: Deploy a multi-node, zero-packet-drop Apache Kafka streaming bus optimized for ingestion rates exceeding millions of network events per second.
Data Lake Serialization: Structure and store the parsed Extended Data Records (XDRs) into a high-capacity time-series database (ClickHouse or Apache Iceberg) optimized for deep analytics.
3. AI/ML Pipeline Integration (AIOps Ready)
Feature Store Architecture: Expose low-latency streaming endpoints (via Kafka/gRPC) formatting raw network metrics into structured data streams for downstream AI/ML anomaly detection models.
In-Stream Anomaly Logic: Implement basic stream-processing rules (using Apache Flink or ksqlDB) to flag instant micro-bursts and latency spikes before they hit user-facing layers.
4. Visualizations & Business Intelligence
Executive & Ops Dashboards: Build Grafana or OpenSearch dashboards translating technical parameters (p99 HTTP/2 SBA latencies, GTP packet drop rates, BNG saturation) into clear business outcomes (SLA status, Customer Experience Scores).
Telco-Grade RBAC: Implement Role-Based Access Control separating low-level infrastructure views from executive reporting layers.
Required Freelancer Skills
eBPF / Kernel Tracing: Expert in C/Go, eBPF maps, ring buffers, and the Linux kernel verifier.
Telecom Networks: Thorough understanding of 3GPP 5G Standalone architectures, SBA, and Wireline broadband routing (BNG, BRAS, PPPoE).
Data Engineering: Advanced mastery of Apache Kafka, stream processing, and big-data analytics platforms (ClickHouse, Snowflake, or Thanos/Prometheus).
DevOps: Hands-on experience with Kubernetes, OpenShift, Helm, and cloud-native infrastructure automation.
Instead of heavy legacy sidecars, the solution must use eBPF (extended Berkeley Packet Filter) for host/container instrumentation, routing high-capacity XDR streams into an AI-ready data lake.
Detailed Scope of Work & Requirements
1. Infrastructure & Probe Deployment (Cloud-Native)
Kubernetes/NFV Instrumentation: Create IaC (Helm/Terraform) to deploy eBPF collection agents across multi-tenant Kubernetes, Red Hat OpenShift, and legacy NFV nodes.
Kernel Integration: Configure kprobes and uprobes to trace traffic safely inside the host Linux kernel, targeting 5G Network Slices and core workloads without exceeding 1.5% host CPU overhead.
High Availability: Design the agents for a 24/7 mission-critical operational environment with automated self-healing loops.
2. Protocol Decoding Engine & High-Capacity Pipeline
Protocol Decoding: Build custom parsing layers within the data pipeline to decode Mobile (GTP-U, PFCP, HTTP/2, SBA) and Wireline (PPPoE, DHCP, QoS, BNG/BRAS) protocols.
Kafka Streaming Tier: Deploy a multi-node, zero-packet-drop Apache Kafka streaming bus optimized for ingestion rates exceeding millions of network events per second.
Data Lake Serialization: Structure and store the parsed Extended Data Records (XDRs) into a high-capacity time-series database (ClickHouse or Apache Iceberg) optimized for deep analytics.
3. AI/ML Pipeline Integration (AIOps Ready)
Feature Store Architecture: Expose low-latency streaming endpoints (via Kafka/gRPC) formatting raw network metrics into structured data streams for downstream AI/ML anomaly detection models.
In-Stream Anomaly Logic: Implement basic stream-processing rules (using Apache Flink or ksqlDB) to flag instant micro-bursts and latency spikes before they hit user-facing layers.
4. Visualizations & Business Intelligence
Executive & Ops Dashboards: Build Grafana or OpenSearch dashboards translating technical parameters (p99 HTTP/2 SBA latencies, GTP packet drop rates, BNG saturation) into clear business outcomes (SLA status, Customer Experience Scores).
Telco-Grade RBAC: Implement Role-Based Access Control separating low-level infrastructure views from executive reporting layers.
Required Freelancer Skills
eBPF / Kernel Tracing: Expert in C/Go, eBPF maps, ring buffers, and the Linux kernel verifier.
Telecom Networks: Thorough understanding of 3GPP 5G Standalone architectures, SBA, and Wireline broadband routing (BNG, BRAS, PPPoE).
Data Engineering: Advanced mastery of Apache Kafka, stream processing, and big-data analytics platforms (ClickHouse, Snowflake, or Thanos/Prometheus).
DevOps: Hands-on experience with Kubernetes, OpenShift, Helm, and cloud-native infrastructure automation.