Your points are good. I don't know a good macrobenchmark at present, but at least various latency numbers are easy to get out of fio. I ran a similar set of tests on an Optane 900P with results below. 'clat' is, as fio reports, the completion latency, measured in usec. 'configuration' is [block size], [iodepth], [jobs]; picked to be a varied selection that obtained excellent throughput from the drive. Table reports average, and 99th percentile, latency times as well as throughput. It matches Ignat's report that large block sizes using the new option can have worse latency and throughput on top-end drives, although that result doesn't make any sense to me. Happy to run some more there or elsewhere if there are suggestions. devicetype configuration MB/s clat mean clat 99% ------------------------------------------------------------------ nvme base 1m,32,4 2259 59280 67634 crypt default 1m,32,4 2267 59050 182000 crypt no_w_wq 1m,32,4 1758 73954.54 84411 nvme base 64k,1024,1 2273 29500.92 30540 crypt default 64k,1024,1 2167 29518.89 50594 crypt no_w_wq 64k,1024,1 2056 31090.23 31327 nvme base 4k,128,4 2159 924.57 1106 crypt default 4k,128,4 1256 1663.67 3294 crypt no_w_wq 4k,128,4 1703 1165.69 1319 Ignat, how do these numbers match up to what you've been seeing? -John On Tue, Aug 18, 2020 at 5:23 PM Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote: > > On Tue, Aug 18, 2020 at 2:12 PM Ignat Korchagin <ignat@xxxxxxxxxxxxxx> wrote: > > > > Additionally if one cares about latency > > I think everybody really deep down cares about latency, they just > don't always know it, and the benchmarks are very seldom about it > because it's so much harder to measure. > > > they will not use HDDs for the workflow and HDDs have much higher IO latency than CPU scheduling. > > I think by now we can just say that anybody who uses HDD's don't care > about performance as a primary issue. > > I don't think they are really interesting as a benchmark target - at > least from the standpoint of what the kernel should optimize for. > > People have HDD's for legacy reasons or because they care much more > about capacity than performance. Why should _we_ then worry about > performance that the user doesn't worry about? > > I'm not saying we should penalize HDD's, but I don't think they are > things we should primarily care deeply about any more. > > Linus >
