I want to replace Prometheus for long-term storage, what should I use?

If you want to replace Prometheus for long-term storage, the most common answer is: don’t replace Prometheus entirely—add a long-term remote storage backend.

Best options

1. Grafana Mimir

• Probably the most direct “Prometheus at scale” replacement for long-term metrics storage
• Compatible with Prometheus remote_write
• Designed for high availability, horizontal scaling, and multi-tenancy
• Good choice if you want to keep using Prometheus-style scraping and PromQL

2. Cortex

• Older but still used
• Similar goals to Mimir
• Also compatible with Prometheus remote_write
• If starting fresh, Mimir is usually preferred over Cortex

3. Thanos

• More of an extension to Prometheus than a replacement
• Uses Prometheus + sidecars + object storage
• Great for:
  • long-term retention
  • global querying across many Prometheus servers
  • downsampling
• Good if you already run Prometheus and want to keep it

4. VictoriaMetrics

• Very popular for long-term metrics storage
• Efficient, simpler operationally than Mimir/Cortex in many cases
• Supports Prometheus ingestion via remote_write
• Good performance and lower resource usage
• Often a strong choice if you want something practical and cost-effective

---

Quick recommendation

• Want the most Prometheus-like scalable platform? → Grafana Mimir
• Want simplest long-term metrics storage with good performance? → VictoriaMetrics
• Already have Prometheus and just want long-term retention? → Thanos

---

If you mean “replace Prometheus entirely”

Prometheus itself is usually still used for:

• scraping targets
• alerting
• recording rules

The long-term storage layer is typically separate.

If you truly want to stop using Prometheus servers, then a system like Mimir or VictoriaMetrics can ingest metrics directly via remote_write from agents like:

• Prometheus
• Grafana Agent / Grafana Alloy
• vmagent

---

My practical suggestion

For most teams:

• VictoriaMetrics if you want simpler ops and strong efficiency
• Mimir if you’re already in the Grafana ecosystem and expect larger scale
• Thanos if you want to keep your current Prometheus architecture

If you tell me:

• your scale
• cloud/on-prem
• retention period
• whether you need alerting/querying in PromQL

…I can recommend one specifically.