To distribute a file, OneFS reduces it to blocks of data, arranges the blocks into stripe units, and then allocates the stripe units to nodes over the internal network.
At the same time, OneFS distributes erasure codes that protect the file. The erasure codes encode the file's data in a distributed set of symbols, adding space-efficient redundancy. With only a part of the symbol set, OneFS can recover the original file data.
Taken together, the data and its redundancy form a protection group for a region of file data. OneFS places the protection groups on different drives on different nodes—creating data stripes.
Because OneFS stripes data across nodes that work together as peers, a user connecting to any node can take advantage of the entire cluster's performance.
By default, OneFS optimizes striping for concurrent access. If your dominant access pattern is streaming—that is, lower concurrency, higher single-stream workloads, such as with video—you can change how OneFS lays out data to increase sequential-read performance. To better handle streaming access, OneFS stripes data across more drives. Streaming is most effective on clusters or subpools serving large files.