Case Study: Primer.ai achieves sub-millisecond query performance and improved ROI with StarTree Cloud

Primer.ai cuts MTTR to under 60 seconds with StarTree

Primer.ai, an AI-powered software company specializing in real-time intelligence analytics, faced significant scaling challenges with their open-source Apache Pinot deployment. As their data volume grew, high-frequency Upsert operations caused severe latency spikes and placed a heavy strain on their internal engineering teams. To address this, they partnered with vendor StarTree to move their operations to StarTree Cloud, a professionally managed implementation of Apache Pinot.

StarTree implemented a comprehensive solution that consolidated Kafka partitions, moved Upsert metadata off-heap, and enforced strict data colocation for zero-shuffle joins. This resulted in stabilized p99 tail latencies, eliminated query timeouts, and restored sub-millisecond query performance. The move to StarTree Cloud also improved operational ROI by replacing self-managed infrastructure with a fully managed platform, reducing their Mean Time to Recovery from nearly an hour to under 60 seconds.

Primer.ai