What changes when your search index lives on S3?
Summary
Infino explores what happens when a search index is stored in object storage (S3, GCS, Azure Blob). The article discusses architecture changes (superfiles and manifests), a cost model driven by network round trips, and read strategies (lazy, range-based reads) that optimize performance when the index lives off-disk. It also covers caching, indexing structures (manifest, term index), and notes that core search algorithms remain similar to Lucene, with storage as the primary driver of performance and cost.