Guide · April 2025
Managed observability
without lock-in.
Most teams pick "open source" when they really want "not locked in." Those are different problems. Here’s how open standards give you data portability without forcing you to run the stack yourself.
The open-source trap.
Engineering teams default to "let’s self-host the open-source version" because they remember the Datadog bill. The fear is real. The assumption — that open source is the only way to stay portable — is not.
Running SigNoz, Grafana Stack, or OneUptime yourself is a meaningful ops job. Someone on your team becomes the unofficial ClickHouse operator. Storage fills up. Retention debates happen. Upgrades break things. A platform team of two spends 20% of their quarter keeping "free" observability running.
What most teams actually want is: predictable pricing, no vendor trap, and not having to babysit the stack. You can get all three — but the answer isn’t open source. It’s open standards.
Lock-in is a data problem, not a licensing problem.
When you migrate off Datadog, the hard part isn’t cancelling the contract. It’s that your 18 months of historical traces, logs, and metrics are in a proprietary format you can’t extract. You rebuild dashboards, alerts, and on-call policies from scratch. That cost is the lock-in — whether the vendor is "open source" or not.
Open standards solve this:
- —Ingest: OpenTelemetry (OTLP). The industry-standard wire protocol. Point your OTLP exporter at any OTLP-compatible backend — SaviourOps, SigNoz, Tempo, Datadog — and keep moving. No SDK rewrites.
- —Storage: ClickHouse (open). The storage format is the same one used by SigNoz, OneUptime, and Uber’s internal stack. Your data structure is readable by any ClickHouse client.
- —Export: OTLP + Parquet + CH native. One-click export of your full history in standard formats. Load it into self-hosted ClickHouse, BigQuery, Snowflake, or S3 for long-term archival. You own your data.
The four approaches compared.
| Approach | Ingest | Storage | Export | Ops cost | Data portable | Lock-in |
|---|---|---|---|---|---|---|
| SaviourOps (managed + open standards) | OTLP, eBPF | ClickHouse (open) | OTLP, Parquet, CH native | Zero — fully managed | ✓ | None — export anytime |
| SigNoz / OneUptime / OSS self-host | OTLP | ClickHouse (open) | ClickHouse queries | High — you run it | ✓ | None, but ops cost is real |
| Grafana Cloud | OTLP, Prometheus | Loki/Mimir/Tempo (proprietary in cloud) | Queries only | Low — managed | Partial | Medium — re-ingesting elsewhere is hard |
| Datadog / New Relic / Honeycomb | Proprietary + OTLP | Proprietary | Read API only | Low — managed | — | High — leaving means starting over |
The real cost of "free."
A typical Series A team self-hosting SigNoz or OneUptime on EKS:
- —3–5 nodes for the data plane: ~$800–1,500/mo AWS
- —Storage (100 GB/mo, 30-day retention): ~$300–500/mo
- —Engineer time: ~4–8 hrs/week triaging cluster issues, upgrades, capacity = ~$2,000–4,000/mo loaded cost
- —Opportunity cost: features not shipped because someone was firefighting the o11y stack
Total true cost: $3,000–6,000/month, well above what a managed SaaS would charge for the same load. "Free" isn’t free — it’s a cost you pay in team attention.
When self-hosting is actually right.
Self-hosting OSS makes sense in a few specific cases:
- —Regulatory environments where data residency is legally binding (HIPAA, PCI, FedRAMP) andmanaged BYOC isn’t offered.
- —Very large scale (500+ engineers, 10+ TB/day ingestion) where managed markup would exceed the cost of running it.
- —Platform teams that already run ClickHouse and Kubernetes at scale and want observability as an extension of existing operational muscle.
If you’re not in one of those categories, "managed but portable" is the better path. That’s what we built SaviourOps to be.
How SaviourOps is built for portability.
- —OTLP-native ingestion. Every data path accepts OTLP. No proprietary SDKs required.
- —ClickHouse storage. The same open-source database you could run yourself. Your schema is readable with any ClickHouse client.
- —One-click data export. Export full history in OTLP, Parquet, or ClickHouse native format anytime. We even ship a migration guide for moving to self-hosted SigNoz.
- —Enterprise BYOC. For strict data-residency needs, we run the data plane inside your AWS/GCP/Azure account. Same product, your cloud.
- —Transparent pricing. $0.30 per GB ingested. Free tier (3 GB/mo) forever. No surprise overages — caps and warnings built in.
Your next incident is coming.
Answer it in seconds.
No dashboards to stitch together. No PagerDuty invoice. Deploy in minutes and stop dreading the pager.
Free tier available. No credit card. No sales calls unless you want one.