On Mon, Mar 27, 2023 at 02:06:09PM -0700, Bart Van Assche wrote: > Using REQ_OP_ZONE_APPEND may lead to reordering - data being written on the > storage medium in another order than the order in which the > REQ_OP_ZONE_APPEND commands were submitted. Only if that reordering actually happens. Which we're very good at avoiding. e.g. when looking at btrfs on ZNS drives we see that this reordering does not actually happen to a quantifiable amount. > Hence, the number of extents > for large files increases and performance when reading large files reduces. > To me comparing the performance of these two approaches sounds like a good > topic for a research paper. I'm not sure that REQ_OP_ZONE_APPEND is better > for all zoned storage workloads than REQ_OP_WRITE. For REQ_OP_WRITE you absolutely must avoid reordering, so you need to globally serialize. If you can come up with a workload where your write based approach is fast, please show it!
