Angrui Technology migrated from RDS MySQL to MatrixOne Cloud Serverless, reducing operational costs by nearly 50% with order-of-magnitude performance improvements.
UNRE is a high-tech company applying opto-mechatronic integration, AI and algorithms to construction digitalization, building laser 3D scanning equipment, intelligent modelling systems and automated BIM generation. Its UCL360 as-built measurement robot produces structured measurement data automatically, and UNRE serves more than 500 enterprises.
Construction quality suffers from the way measurement is done: multiple parties, disparate tools, low efficiency, poor accuracy, delayed and fragmented data, and human error. UNRE's robots solved the capture problem-but the SaaS platform behind them hit database limits. Running on hosted MySQL, single tables passed 100 million rows and complex report queries slowed to a crawl. Data upload and dashboard analysis shared one instance, so the two workloads interfered with each other. And with full backups plus binary logs, backup consistency was complicated to manage and recovery took a long time.
UNRE rebuilt the platform on OmniFabric Cloud serverless instances. One shared database instance became one instance per customer, cutting single-table volume by several orders of magnitude. Upload and analysis were split across two instances connected by zero-copy data publish/subscribe, isolating the workloads and ensuring analysis code cannot modify source data. Snapshot backups with point-in-time recovery replaced the full-backup-plus-binlog scheme, completing in seconds with minimal load on the running system.
Data upload and report rendering both improved by an order of magnitude once tenants were split into their own instances and single-table volume dropped from over 100 million rows.
Billing by SQL executed replaced permanently provisioned large instances, cutting operating cost by close to half.
Zero-copy publish/subscribe puts the two workloads on separate instances-heavy dashboard queries no longer slow ingestion, and analysis code cannot accidentally modify source data.