Our backend architecture for Qchex and Decidalo relies on MariaDB as the primary relational store backing both our C# .NET APIs and our asynchronous Python workers. I interact with our MariaDB schemas every day through the Database tool window in PyCharm and via Entity Framework Core and SQLAlchemy in our application code. Two capabilities that save me enormous effort during sprint development are System Versioned Tables and native JSON validation functions. Whenever regional compliance rules on our DPU UK and CEE Apps require an immutable audit trail of consultant billing edits, I simply enable system versioning on the InnoDB table and query historical row states using standard AS OF TIMESTAMP syntax without maintaining messy trigger tables. More recently, we also started using the built in MariaDB Vector search indexing on version 11.4 so our internal Model Context Protocol agents can run semantic similarity lookups right alongside live transactional records in the exact same database instance. Review collected by and hosted on G2.com.

Complex query optimization across hybrid tables is the one area where I occasionally have to step in and tune execution plans by hand. When a reporting query joins a large relational table with values extracted via JSON VALUE from semi structured vendor payloads, the cost based optimizer sometimes underestimates cardinality and chooses a full table scan over a composite index. During a recent performance investigation where Sentry flagged slow database spans on our Qchex endpoints, I had to run EXPLAIN ANALYZE in my SQL console and add explicit FORCE INDEX hints plus generated virtual columns to get sub second response times back. Additionally, configuring MariaDB MaxScale read write splitting rules and testing Galera Cluster node failover in our staging environment required careful timeout tuning so our .NET connection pools would not drop active transactions during a primary node switchover. Review collected by and hosted on G2.com.