reductstore

ReductStore is a data backbone crafted for robots and industrial IoT. It stores images, telemetry, logs and other sensor streams directly on the edge device, pushes the parts that matter to the cloud, and lets you read the collected data with SQL.

The core is open source and released under Apache 2.0, so you can deploy it on the machine without licensing barriers. The runtime is lightweight, so it runs on resource‑constrained edge hardware. Developers access it through a multi‑language SDK, meaning you can write ingestion, query or replication code in the language that matches your platform. Ingested data comes in many forms: images, video, LiDAR point clouds, IMU samples, logs, generic files, ROS bags, and other binary blobs. Each record is stamped with a timestamp, giving the tool a time‑index that you can query. When you read data, you specify a time range and optionally filter by record type, and the backend returns the matching blobs and metadata in SQL‑style rows.

When a robot collects data, ReductStore can batch many small records into a single S3‑compatible object. This reduces the number of API calls and cuts storage costs compared to other object stores. On the other side, you set up a replication path to send the most critical records to the cloud. The system tolerates intermittent connectivity; it keeps the data locally until a network connection can forward the objects, so loss of data is minimized on unreliable links.

Developers in the robotics field appreciate that ReductStore can ingest and serve billions of records while maintaining high throughput. Benchmarks on the website claim 10× faster writes for 100‑KB records than TimescaleDB, 15× faster reads than MinIO, and a 35% lower storage cost than Foxglove. The platform also offers a 99.99 % uptime SLA across production deployments. Projects that build autonomous vehicles, warehouse robots or connected industrial tools tend to use it because those domains produce high‑frequency, unstructured data that few generic databases can handle efficiently. The tool is less suited for traditional relational workloads or small‑scale sensor setups that do not need time‑indexed querying.

Key features

• Time‑indexed storage for multimodal data: images, video, LiDAR, IMU, logs, files, ROS bags and more
• Lightweight runtime that runs on edge devices with limited resources
• Multi‑language SDK for ingestion, querying and replication scripts