Details
-
Improvement
-
Resolution: Unresolved
-
Medium
-
None
-
None
-
None
-
3
-
9223372036854775807
Description
This ticket is to track the semantics of how the large folios are implemented.
Do they impose a minimum "chunk" size for read-modify-write? Does the stripe_size of a file affect the folio size? Or are the folios opportunistically sized (e.g. application read/write size up to the stripe size). Should there be a llite.*.large_folio=N[KMG] tunable that allows specify the minimum folio size for the file, or possibly in the near future the compression chunk size should set this internally, so that the page cache handles the complexity of aggregating multiple pages into a chunk to avoid issues with mmap.
