On 7/29/25 00:23, Martin K. Petersen wrote: >> b) io_opt, size to ??? >> 4) For raid0/raid10/rai5, this value is set to mininal IO size to get >> best performance. > For software RAID I am not sure how much this really matters in a modern > context. It certainly did 25 years ago when we benchmarked things for > XFS. Full stripe writes were a big improvement with both software and > hardware RAID. But how much this matters today, I am not sure. > FWIW, I just posted a patch that aligns writes to stripe boundaries using io_opt: https://lore.kernel.org/all/55deda1d-967d-4d68-a9ba-4d5139374a37@xxxxxxxxxxxxxxx/ I get about a 2.3% performance improvement with md-raid6, but I have an out-of-tree RAID driver that gets more like 4x improvement. If io_opt means different things to different code, might we consider adding another field to the queue limits to give explicit stripe parameters? Tony Battersby Cybernetics
