GreptimeDB

GreptimeDB is an open-source observability database platform that serves as a unified alternative to Prometheus, Loki, and Elasticsearch. It treats metrics, logs, and traces as a single data model of timestamped wide events rather than separate pillars. The software is deployed as a server and can run in a standalone binary for small deployments or a distributed mode for larger scales.

Users employ GreptimeDB as a backend for OpenTelemetry to consolidate observability data into one engine. It supports querying via SQL and PromQL and is designed to handle high-cardinality data and long-term storage. The system is available as a Docker container and provides a web UI for data management, as well as a dedicated Grafana data source plugin for visualization.

Key features

• Unified storage for metrics, logs, and traces in one engine
• Support for both SQL and PromQL query languages
• Native integration with OpenTelemetry, Prometheus remote write, and Jaeger
• Compute-storage separation allowing independent scaling of read replicas
• Primary storage on object stores including S3, GCS, and Azure Blob
• Columnar engine written in Rust with fulltext and inverted indexes
• Tiered caching using memory and local disk for fast writes and queries
• Support for GenAI telemetry via OTel GenAI conventions

The architecture consists of four main components: a stateless Frontend for protocol entry, Datanodes for regional engine processing and persistence, a Metasrv for metadata and routing, and an optional Flownode for continuous flow computation. It utilizes Apache Arrow for its memory model, Apache Parquet for file storage, and Apache DataFusion for its query engine. This design is intended for SREs and developers who need to scale observability beyond the limits of Prometheus or reduce the operational overhead of managing multiple database stacks. The distributed engine allows for horizontal scaling without the need for resharding or manual data migration.

GreptimeDB positions itself as a cloud-native observability backend that reduces storage costs by leveraging object storage while maintaining sub-second query performance across petabyte-scale data.