Vector

Vector is an end-to-end observability data pipeline that puts you, rather than a vendor, in charge of where telemetry goes. It collects, transforms and routes logs and metrics to the destinations you use today and the ones you may move to next year, which is what makes switching vendors survivable.

It is written in Rust, and reliability is the stated primary design goal. The same tool deploys as an agent on each host or as an aggregator for a fleet, so one binary covers the whole path. Logs and metrics share a single data model, with traces listed as coming soon.

The project publishes its own test harness, and the results it reports span several dimensions.

• Throughput: published tests place Vector ahead of Filebeat, Fluentd and Logstash on TCP and file transfer cases.
• Correctness: harness cases cover disk buffer persistence, file rotation, file truncation and SIGHUP handling.
• Memory safety: the Rust implementation brings memory safety, multi-core execution and delivery guarantees together.
• Vendor neutrality: it is open source and tied to no single destination, which is the point of running it at all.
• Cost control: teams use it to cut observability spend, enrich data in flight and consolidate overlapping agents.

Vector is maintained by Datadog's community open source engineering team. It is downloaded more than 100,000 times a day, its largest user processes over 500TB daily, and container images, packaged releases and a quickstart guide cover the ways in.