eXtremeDB 9.0: On-Device AI Search and a Database That Fits in 8 KB
Embedded and edge teams are being asked to do more on the same constrained hardware: run AI inference locally, manage structured sensor data, and still deliver deterministic, reliable behaviour. Our partner McObject's latest release, eXtremeDB 9.0, takes direct aim at that challenge with two headline additions that change what is possible at the edge.
On-device AI: native ANN indexes
eXtremeDB 9.0 introduces native Approximate Nearest Neighbor (ANN) indexes inside the database engine itself. Developers can now run high-speed similarity searches on high-dimensional data without adding a separate library or external service.
Use cases this unlocks:
- Sensor fusion and anomaly detection
- Predictive maintenance and industrial AI
- Computer vision and pattern recognition
- Edge-deployed machine-learning inference
- Recommendation and personalisation on-device
Because the indexes are native and transactionally consistent, you get AI-grade search that behaves predictably in production, not just in the lab. Supported environments include AMP architectures, multi-core CPUs and resource-constrained systems.
A database that fits in 8 KB: the Edge Client
The new eXtremeDB Edge Client is built for the smallest IoT devices. It requires 8 KB of memory or less, enabling devices that could never host a database before to handle structured, transactional data.
It works as a write-first store:
- Uses the same schema definitions (DDL) as full eXtremeDB
- Supports start, commit and abort transactions
- On commit, sends data upstream via McObject's Active Replication Fabric, then reclaims local storage
- Preserves referential integrity for multi-sensor and multi-record devices before transmitting
The data arriving upstream is already clean and consistent.
What else is new in 9.0
Version 9.0 also delivers improved MVCC transaction scheduling, performance work on the Python, Java and C# APIs, .NET 8 support (while retaining .NET Framework compatibility), Python up to 3.14, persistent kernel runtime parameters, richer time-series analytics, expanded SSL/TLS options including wolfSSL and Mbed TLS, and a wider REST API.
Why it matters, and how Logic Technology helps
For teams building intelligent products on constrained hardware, this is a meaningful step: AI-grade search and structured data handling, without giving up the small footprint and deterministic performance embedded systems demand.
As McObject's Benelux partner, Logic Technology helps you decide where eXtremeDB fits in your architecture and guides you from evaluation through integration. Data management is one of our core areas, and whatever your platform, there is always a Logic solution.
