GreptimeDB leads the shift toward unified Rust-native observability

Teams are replacing fragmented observability stacks with GreptimeDB to unify metrics, logs, and traces in a single Rust-native engine. Most organizations currently manage three separate databases for metrics, logs, and traces, which leads to three separate bills and three different tools to break. GreptimeDB uses a codebase that is 99.3% Rust, sitting on top of Apache Arrow and DataFusion. This architecture allows the database to replace the Prometheus, Loki, and Elasticsearch trio. In TSBS benchmarks simulating CPU metrics from 4,000 servers over three days, GreptimeDB achieves 2.17x the write throughput of TimescaleDB, recording 285,301 rows per second against 131,531 rows per second. The test data includes 103.6 million rows containing 10 tags like hostname and region, along with 10 metric fields. It also outperforms TimescaleDB in 13 of 15 query types. It provides a 62x speedup for single-host, long time-range queries and a 67x speedup for the cpu-max-all-8 query. OceanBase Cloud manages 300 TB of logs across 80 clusters using this technology.

Massive efficiency gains in storage and query execution

The shift toward GreptimeDB stems from massive efficiency gains in storage and query execution. GreptimeDB utilizes a columnar format and a vectorized query engine built on Apache DataFusion to process data in batches, which allows it to outperform TimescaleDB in 13 of 15 query types during standard TSBS testing. The Mito engine uses an LSM-Tree architecture optimized for append-only patterns, while the storage layer uses Apache Parquet to group similar data. This configuration results in a storage footprint that is only 1/18 the size of TimescaleDB, which used 20 GB for the same 103.6 million rows compared to the 1.1 GB used by GreptimeDB. Because it supports object storage like Amazon S3, Google Cloud Storage, and MinIO, teams reduce storage costs by up to 50 percent. While GreptimeDB leads in heavy aggregations, TimescaleDB remains 8.7x faster for lastpoint queries that find the latest value for a specific host. InfluxDB 3 also uses a columnar architecture, but its July 2026 release of the Pacha Tree engine uses a proprietary .ptfiles format that remains unreadable by earlier versions. You should verify if your existing InfluxDB clusters can handle a migration to these new formats before you commit to an upgrade.

Performance and resource management roadmap

Current adoption trends follow the GreptimeDB roadmap, which focuses on performance and resource management throughout 2026. The v1.2.0 release in June 2026 introduced a new structural stored JSON data type called JSON2 and supports Prometheus Remote Write v2 ingestion. This release also added online partitioning and experimental flow incremental reads. In September 2026, the v1.3 release adds metric engine optimization, compaction optimization, and vector index and AI functions. The roadmap also includes a two-phase adaptive resource management project slated for completion in the fourth quarter. This later phase includes remote compaction and indexing capabilities that are production ready. Furthermore, the v1.5 release in late 2026 adds Pandas DataFrame SQL query support. Li Auto uses the system for telemetry across its electric vehicle fleet, and OceanBase Cloud relies on it for large-scale log management. GreptimeDB provides Apache 2.0 licensing for its core, although users must pay for enterprise features like soft-drop and recovery. Does the growing list of specialized features meet the needs of teams moving away from the mature PostgreSQL ecosystem?