The idea to create a new timeseries database came from our experience in development of AI applications for industrial production processes. We couldn’t use classical TSDBs because they focused on measurement from sensors, but we had large images from computer vision cameras and 48kHz sound as well. We needed a format agnostic blob storage for our data zoo, something like S3 but with an API which provided access with timestamps and time intervals. Moreover, the disk space on our edge devices was critical to us and data reduction was a must.
In December 2021, Alexey Timin launched the Reduct Storage project and started developing a prototype of a storage engine in modern C++20.
The prototype was ready in 5 months after the project started, and we had a chance to test it with a real AI application. After tons of bug fixes we got a working engine which did everything what we needed with good performance.
Meanwhile, Ciarán Moyne joined the project to support it with Client SDKs and CLI tools. Together we wrote a motivation article “Why we need a new database for Industry 4.0 to tell different communities about the project and get feedback. We didn’t blow up the Internet of course, but people who work in the AI branch were positive about the general idea.
The open source community was also growing. We started receiving bug fixes, documentation improvements and questions.
You can use the experimental implementation (v0) for any purposes, but it doesn’t have many enterprise features like user authorisation or replication. We’re planning to add them in the next major version. See our roadmap