On Thu, Apr 15, 2021 at 04:41:06PM +0100, John Garry wrote: > On 15/04/2021 13:18, Ming Lei wrote: > > On Thu, Apr 15, 2021 at 11:41:52AM +0100, John Garry wrote: > > > Hi Ming, > > > > > > I'll have a look. > > > > > > BTW, are you intentionally using scsi_debug over null_blk? null_blk supports > > > shared sbitmap as well, and performance figures there are generally higher > > > than scsi_debug for similar fio settings. > > I use both, but scsi_debug can cover scsi stack test. > > > > Hi Ming, > > I can't seem to recreate your same issue. Are you mainline defconfig, or a > special disto config? The config is rhel8 config. As I mentioned, with deadline, IOPS drop is observed on one hardware(ibm-x3850x6) which is exactly the machine Yanhui reported the cpu utilization issue. On another machine(HP DL380G10), still 32cores, dual numa nodes, IOPS drop can't be observed, but cpu utilization difference is still obserable. I use scsi_debug just because it is hard to run the virt workloads on that machine. And the reported issue is on megaraid_sas, which is a scsi device, so null_blk isn't good to simulate here because it can't cover scsi stack. Thanks, Ming
