Network traffic visualization — future enhancement ideas

[nadaverell]

This is a collection of ideas for improving network traffic visualization in Radar. These are not planned or prioritized yet — just a place to track possibilities and gather community input.

Feedback and additional ideas welcome!

---

1. Surface L7 detail in the traffic graph (Hubble)

Hubble already streams HTTP method, path, status code, and gRPC protocol per flow. This data reaches the backend but never makes it to the UI — edges currently only show connection counts and bytes.

Idea: when Hubble is the active source, clicking an edge could show top endpoints (GET /api/health → 200, POST /checkout → 500), HTTP status breakdown, and error rate. This is the clearest differentiator between Radar and a generic Grafana dashboard.

---

2. Pixie as a traffic source

Pixie (CNCF) is an eBPF-based observability tool that goes beyond "who talks to whom" — it captures actual request payloads, SQL queries, and latency histograms. It fills a niche none of Radar's current sources (Hubble, Caretta, Istio) cover: answering why a specific service is slow.

Idea: add Pixie as a 4th traffic source, detected via the presence of vizier deployments.

---

3. Linkerd as a traffic source

Linkerd has significant market share, especially among teams who found Istio too heavy. It exposes Prometheus metrics similar to Istio. Since Radar already has Prometheus query infrastructure from the Caretta/Istio sources, this should be relatively low effort.

---

4. Istio configuration health (Kiali-lite)

Radar already parses Istio VirtualService and DestinationRule CRDs for topology edges. The next step could be enriching the Istio resource drawers with config-level insight: traffic weight splits, mTLS policy status, broken VirtualService detection. All readable directly from CRD specs — no additional data source needed.

---

5. Azure Retina as a traffic source

Retina is Microsoft's eBPF-based network observability tool for AKS, now a CNCF sandbox project. It exposes Prometheus metrics, so detection and querying would follow the same pattern as Caretta/Istio. Useful for AKS users who get little value from native Azure Monitor maps.