Hello, I am running a few tests (Lustre 2.4.2) doing N -> 1 writes, strided (worst possible performance, I know). Are there some optimizations specifically for N -> 1 writes? N -> N writes is what I expected it to be (compared to another system) thanks, Ron