1 paper
Carl Kugblenu, Petri Vuorimaa
Production vector search systems often fan out each query across parallel lanes (threads, replicas, or shards) to meet latency service-level objectives (SLOs). In practice, these l…