myRestro Manager
.jpeg)
About the Project
An enterprise-level, multi-tenant SAAS platform engineered for modern restaurant chains to orchestrate table reservations, real-time order flows, and live kitchen execution pipelines. Unlike traditional POS setups that require proprietary hardware networks, myRestro leverages a cloud-first, serverless architecture that enables instant table synchronization, automated order queue routing, and detailed analytics dashboards. The application serves as the single source of truth for front-of-house staff, kitchen crews, and management alike.
The Challenge
During peak dinner hours, concurrent updates from hundreds of active tables created database deadlocks and extreme query latency. Kitchen displays would fail to synchronize with orders placed by waiters, leading to duplicate orders or missed items. Our original implementation relied on HTTP polling, which overwhelmed our database connection pool (PostgreSQL) and created unacceptable latency spikes of up to 8 seconds under high concurrent loads.
The Solution
I redesigned the network layer, migrating from polling to persistent duplex WebSockets with automatic backoff reconnection algorithms. I optimized our database access by writing custom Prisma query filters, applying composite indices on table states, and offloading live order queues to an in-memory Redis cache. For front-end rendering, I implemented React virtualized lists and memoized chart states, ensuring smooth rendering performance even as high-volume data points streamed in continuously.
Technologies Used
