What are the methods for long-term storage of metrics in the Prometheus system?

Prometheus by default stores metrics locally in TSDB (Time Series Database) format, but this storage is designed for a relatively short period (usually a few weeks). For long-term storage of metrics, the following approaches are used:

1. Remote Write / Remote Read — Prometheus can send data to external storage systems via the remote_write mechanism. This allows integrating Prometheus with external databases that support long-term storage and scaling.

2. External storage systems (Long-term storage):

   • Thanos — an extension for Prometheus that aggregates data from multiple Prometheus instances and stores it in object storage (S3, GCS, etc.) with long-term storage and query capabilities.
   • Cortex — a distributed metric storage system compatible with Prometheus, supporting long-term storage and scaling.
   • VictoriaMetrics — a high-performance TSDB that can be used as long-term storage.

3. Data export — periodic export of metrics from Prometheus to other analytics systems or databases (e.g., InfluxDB, Elasticsearch).

Thus, for long-term storage of metrics, Prometheus is usually integrated with external systems via remote_write and specialized solutions like Thanos or Cortex.