ReductSoftware UG

In many applications, you may need to capture a history of unstructured data. Previously, we built an in-house solution based on Minio and InfluxDB. Apart from the fact that it wasn't an easy development task, we also ran into performance issues with blobs smaller than 100KB over high latency networks. ReductStore is a time series database for blob data designed specifically for edge computing. It uses HTTP to access data over time intervals and shows much better performance for small blobs due to its batching approach.

ReductStore is a time series database designed to manage large amounts of blob data. For this kind of application, performance is critical. We decided to write it in C++. It was a good decision until we started porting the database to new platforms. Last month I was mostly busy with cross-compilation issues and different behavior on different platforms. We decided to switch to Rust because it has a package manager and cross-compilation out of the box. For a smooth migration we are using cxx.rs library and rewriting modules step by step with deep testing intermediate versions. I guess in a few weeks the migration will be done and ReductStore will be 100% Rust.

ReductStore provides an HTTP API to access data. This allows us to easily integrate it into any cloud environment. In addition, you can scale it by spinning up a few instances of the database and load balancing through NGINX. In this tutorial, you will learn you how you can launch two ReductStore instances by using Docker Compose and split them by using subpaths in NGINX.

We tried using Minio to store a history of images from a CV camera. Unfortunately, it didn't work well because it didn't provide an API to retrieve data over a time interval, and it didn't do data redaction for old data. ReductStore is designed for such a kind of applications and works well with a few terabytes of historical data.