Case Study: iFood achieves 9X database cost savings with ScyllaDB

iFood Relies on ScyllaDB to Deliver Over 100 Million Events a Month to Restaurants

iFood, Brazil’s leading food delivery app, needed a scalable way to deliver and poll over 100 million events a month as its order volume surged and restaurant network expanded. Its legacy PostgreSQL and Apache Ignite-based approaches struggled with joins, failover, monitoring, and bursty traffic, while DynamoDB’s autoscaling and cost profile were not a good fit. iFood turned to ScyllaDB for its Connection-Polling service to support reliable event delivery across AWS-based infrastructure.

With ScyllaDB, iFood implemented a similar NoSQL data model and moved polling workloads to a highly available, low-latency cluster. The results were strong: a 3-node ScyllaDB deployment on c5.2xlarge instances met throughput needs while cutting database costs from about $4,500 per month to $500 per month, a 9x savings. iFood also reduced event handling time from roughly 80 ms with DynamoDB to about 3 ms with ScyllaDB, used one-eighth the infrastructure for writes, and trimmed 40% of the code.

iFood

Thales Biancala

iFood