ReductStore

ReductStore v1.21 shipped on 18 September with a Data Explorer in the Web Console. Conditional queries used to mean writing the JSON by hand. Now you pick a bucket and an entry, set a time range, filter on labels, sample or limit the result, and reorder the processing stages visually. ReductROS and ReductSelect steps are configured the same way. https://lnkd.in/e8qEZy6S #release #datainfrastructure #opensource

🚀 ReductStore v1.21 is out! This release is a special one for us — not only because of the new features, but because of how much of it was shaped by our growing open-source community. 🔎 New Data Explorer Build conditional queries visually, sample and filter data, configure ReductROS and ReductSelect processing steps, and still access the generated JSON when you need it. 🔄 More powerful replication Destination prefixes, gzip/zstd compression, path-aware entry patterns, automatic bucket creation, and improved throughput. 📊 Better system diagnostics Per-bucket usage statistics and server logs are now available directly through the $system bucket. 🌱 More community contributions This release cycle brought together contributors across ReductStore Core, CLI, and the Web Console. A huge thank you to everyone who contributed code, feedback, and ideas. And there’s more: lifecycle improvements, a more consistent query syntax, CLI updates, and changes to authentication configuration. Read the full release notes 👇 https://lnkd.in/dp9h_JFG Many thanks to Dibbayajyoti Roy, Malhar Vora, Rohan Kumar Dubey #ReductStore #OpenSource #Robotics #EdgeComputing #IIoT #DataEngineering #Rust #Database

Deleting old data on a schedule usually means a cron job and a script somebody has to keep working. ReductStore v1.20 makes it a policy on the store instead. Lifecycle policies are background tasks that delete records, or compress persisted blocks with zstd. You scope one by entry name, by record age, and by a conditional expression over labels. Every policy runs in enabled, dry_run or disabled mode, so you can see what it would remove before it removes anything. Diagnostics land in the $system/lifecycle entry. https://lnkd.in/ef4_-H_x #IIoT #edge #release

A colleague wants the twenty minutes of data behind a bug report. The usual answer is an API token they should not have, or a zip file you export by hand. Since v1.17 you can create a query link instead. It carries your read permissions, it is scoped to one query, and it expires at a timestamp you set. Whoever holds the link gets those records and nothing else. No token, no account, and it stops working when the expiry passes. https://lnkd.in/eDkrXGgC #datainfrastructure #sharing #IIoT

A rosbag is a file format, not a database. Record a fleet for a month and retention turns into a directory of thousands of bag files and a script that decides which ones to delete. ReductStore writes continuously instead, and retention is a FIFO quota by volume set on the bucket. Records are removed only when the storage limit is reached. So a quiet week costs you nothing, and a busy one does not fill the disk. The script goes away with the files. https://lnkd.in/e7fEj-us #robotics #ROS #rosbag

Here we show where #InfluxDB works well — and where it doesn’t — in robotics applications. You can also see how naturally ReductStore can be added to the stack as an additional blob storage layer for keeping the history of raw sensor data. You don’t need to choose between a #TSDB and ReductStore. Use both. It’s a much simpler approach than trying to manage raw sensor data with #S3-like storage or a filesystem. #robotics #datastorage #ROS

A robot publishes on sensors/imu/raw and cameras/front/jpeg. Most stores make you flatten that into one namespace and rebuild the hierarchy from a naming convention. Since v1.19, ReductStore entry names are paths with multiple segments. So the store is laid out the way the machine is, and the same shape covers ROS topics, Zenoh key expressions and MQTT topics. The same release added ReductBridge, which labels ROS messages on the way in and keeps their schemas and metadata as attachments. https://lnkd.in/engY7AWn #robotics #ROS #release

A Protobuf payload is useless without its descriptor, and the descriptor usually lives somewhere else entirely. Since v1.20 ReductStore stores it beside the data. ReductBridge ingests Protobuf payloads, extracts labels from the message fields, and writes the descriptor to the entry as a $schema attachment. ReductSelect reads that same attachment on the way out and exposes the decoded fields to SQL (DataFusion). One descriptor, written once at ingest. https://lnkd.in/ef4_-H_x #protobuf #sql

An edge device fills its disk on a Sunday and nobody is there to clear it. Recording stops and the week has a hole in it. ReductStore puts the ceiling on the bucket instead. Set the quota type to FIFO and a size in bytes, and the store deletes the oldest block to make room once the limit is reached. The ceiling is volume, not a retention window, so the disk behaves the same whether the sensors are quiet or the robot is running hard. https://lnkd.in/eRf3g5RA #edge #datainfrastructure #IIoT

Robots do not always have a network connection when they need one. Streaming everything is not realistic, and recording nothing is worse. ReductStore and INSAION are now official partners. ReductStore already runs inside the INSAION Agent as the edge storage engine. It writes camera, LiDAR and control data into a local Rolling Buffer that cycles out the oldest data once the disk fills. When an alarm fires or an operator flags something, INSAION's Surgical Cloud Sync queries ReductStore by time and label and uploads just that window. Full fidelity stays on the robot. Only the incident goes over the network. Both teams wrote about how it works, from their side. https://lnkd.in/eY923yhk https://lnkd.in/e-fkk4_6 #robotics #edge #ROS