Repository Architecture & Sizing – managing large-scale datasets (tens to hundreds of terabytes) the performance depends less on the total volume of data and more on the total number of files and how you structure your workflow because restic maintains an index of all data blocks in system memory during operations scalability requires careful planning around RAM pack sizes and backend selection
File count vs. data volume – performance is constrained more by the total number of files than raw Gigabytes budget roughly 1 GB of RAM per 7 million unique files and 1 GB of RAM per 7 TB of data as a rough scaling baseline
Split into multiple repositories- instead of one massive repo divide your data into separate logical repositories (e.g. by service host or data tier) this lowers peak RAM usage during operations like check and prune
Tune pack sizes – for large repositories consider increasing the default pack size using –pack-size (e.g. 50 MB) on initialization/commands to reduce the total number of pack files and lower index memory overhead