Boubacar - I've answered 1) above 2) Snapshot is multiple files. No limitations on snapshot size. But I've only seen users with up to 500GB of disk data, there might be bottlenecks with larger sizes that we haven't tested for yet. 3) Sounds reasonable. In general, the smaller the number of documents, the more performant queries are. 4) Typesense is not meant to be used as your primary datastore. So for disaster recovery, you'd typically want to do regular backups on your primary datastore. That said, you could also snapshot the copy of the data you've sent to Typesense using the snapshot API and start a new cluster with that snapshot. The key thing to consider for a dataset of your size, is the startup time. So it's critical to run multiple nodes in an HA configuration