#030 Vector Search at Scale, Why One Size Doesn't Fit All

Charles Xie, founder of Zilliz, explains how Milvus handles 100B+ vector workloads: a four-tier storage strategy across GPU memory, RAM, local SSD and object storage; a searchable write buffer merged with the main index for real-time search; GPU acceleration at 10k-50k QPS; and where he expects self-learning indices and hybrid dense/sparse search to go. Video: https://www.youtube.com/watch?v=vRlnbRdMZYU