Booking.com selects Weaviate as its next vector database

Booking.com's ML & DS blog says its OpenSearch-backed Embedding Service became harder and costlier to operate as workloads grew. In tests using 100 million embeddings, filtered searches and concurrent reads and writes, the team found dedicated vector databases handled its workloads more efficiently. It selected Weaviate for its consistent performance across the tested scenarios and reported roughly 40% lower usage cost than its OpenSearch baseline at comparable recall and service-level targets.