Every external service you add is another thing that can break. aacyn is one binary — deploy it, and you're done.
Count the boxes. Fewer boxes = fewer failure modes, fewer alerts, less YAML to maintain.
Self-contained. Deploy the binary, get everything.
Single binary — no external database, no time-series store, no query engine. The C columnar store handles ingest, scans, and serving in one process.
Open-source, but requires 3 external services.
ClickHouse for storage, Prometheus for metrics, Grafana for dashboards. Each dependency adds operational overhead and a failure domain.
Great eBPF, but ties you to New Relic.
In-cluster storage handles short-term data. Long-term retention and dashboards require New Relic Cloud or a self-hosted alternative.
eBPF with Grafana's ecosystem requirements.
Beyla emits metrics and traces; you still need Grafana for dashboards and Prometheus or Mimir for storage. Each layer must be deployed and maintained separately.
Powerful, but requires Cilium CNI migration.
Hubble only works on Cilium. Migrating a production CNI is a multi-hour, high-risk operation. If you're not already on Cilium, Hubble is not an option.
Flexible toolkit, but not a unified platform.
Each gadget is a standalone tool. No unified topology view, no built-in golden signals, no pre-configured alerts. Assembly required.