Here is a small, expensive fact about running modern software: to watch one system, most teams run three databases. One for metrics, so you can see that latency spiked at 3 a.m. One for logs, so you can read the error that caused it. And one for traces, so you can follow the request as it bounced through a dozen services on its way to failing. Prometheus, Loki, Elasticsearch - three tools, three query languages, three bills, three things to break.
Greptime, founded in 2022, looked at this arrangement and asked the obvious question that is somehow hard to act on: why three? Its product, GreptimeDB, is an open-source database that stores metrics, logs, traces, and what the industry has started calling "wide events" in a single engine. The pitch is not subtle, and the company does not pretend otherwise - GreptimeDB is marketed as a drop-in replacement for the Prometheus-Loki-Elasticsearch trio. Clear positioning is a courtesy to the buyer. It tells them exactly what to rip out.
The interesting move is architectural. GreptimeDB puts the raw data on object storage - Amazon S3 and its equivalents - rather than on the fast, expensive disks that databases usually demand. Object storage is cheap and effectively infinite; the tradeoff has always been that it is slow. Greptime's bet is that with a columnar format, a careful engine, and compute separated from storage, you can have the cheap storage without the pain. The company claims up to 50x storage cost reduction. That number should be read the way you read all vendor numbers - as the best case, under favorable conditions - but the direction is real, and it is the whole point.
The engine itself is written almost entirely in Rust - about 99.3% of the codebase - on top of Apache Arrow and DataFusion. Rust is having a moment in infrastructure precisely because it offers memory safety without a garbage collector, which matters when you are ingesting a firehose of timestamped events and cannot afford a pause. It also means GreptimeDB is one of the larger Rust database codebases in open source, which is the kind of thing that makes a certain type of engineer star a repository. Enough of them did that GreptimeDB hit number one on GitHub's global trending list within days of its November 2022 launch.
What you can actually do with it is the part that sells. You point OpenTelemetry or Prometheus Remote Write at it, and it ingests. You query it with SQL if you think in tables, or PromQL if you think in time series - both, in the same engine. It derives metrics from raw events on demand rather than forcing you to decide up front what to keep. For a team drowning in observability tooling, the promise is fewer moving parts, a smaller bill, and one place to look when something breaks at 3 a.m.
The founders come by the problem honestly. CEO Xiaodan "Dennis" Zhuang spent 18 years in software, much of it as a senior staff engineer at Alibaba working on distributed messaging, time-series storage, and large-scale observability. His co-founders, CTO Ning Sun and engineering VP Jiachun Feng, carry similar resumes from Alibaba Cloud, Ant Group, and DiDi. These are people who have felt the pain of monitoring hyperscale systems, which is a useful thing to have felt before you try to sell other people a cure.
