Keynote: OpenSearch 3.0 and Lucene 10, Paving the Way for AI and Vector Innovation
summary
OpenSearch 3.0 upgrades to Lucene 10 and a newer JDK, yielding an average latency drop from 160 ms to 16 ms across core queries (about 10× faster). GPU-accelerated vector indexing with NVIDIA Kagra speeds index builds over 9× and cuts costs nearly 4×, moving billion-scale builds from days to hours via a decoupled GPU index-build service. Adds Kafka poll ingestion and gRPC client work from Uber, segment replication improvements from ByteDance, FIPS from SAP, J vector engine from Data Stacks, and Zstd plus SIMD work from Intel.
AI-written summary · about
related talks
26mOpenSearch and the Path to a Trillion-s...
24mVector Search at Uber
48mUnlocking Next-gen Semantic Search With OpenSearch 3.x
43mOpenSearch and the Path to a Trillion-scale Vector Database
33mBeyond Benchmarks: Real-World AI Search Performance in OpenSearch (Lucene Vs Faiss Vs JVector)
41mApache Solr 10: What's Coming up for Vector Search
40mApache Lucene: From Text Indexing to Artificial Intelligence
32mRetrieval Augmentation and Semantic Search at Scale