Date: Fri, 16 Jan 2015 12:13:17 +0300 From: Slawa Olhovchenkov <slw@zxy.spb.ru> To: Mihai Vintila <unixro@gmail.com> Cc: freebsd-stable@freebsd.org Subject: Re: Poor performance on Intel P3600 NVME driver Message-ID: <20150116091317.GA83755@zxy.spb.ru> In-Reply-To: <54B8CEF4.9040206@gmail.com> References: <54B7F769.40605@gmail.com> <20150115175927.GA19071@zxy.spb.ru> <54B8C7E9.3030602@gmail.com> <54B8CEF4.9040206@gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Jan 16, 2015 at 10:42:28AM +0200, Mihai Vintila wrote: > Actually seems that i've was wrong vfs.zfs.top_maxinflight does not have > any influence and this is confirmed by the nvmecontrol perftests. > nvmecontrol perftest -n 32 -o read -s 512 -t30 nvme0ns1 > Threads: 32 Size: 512 READ Time: 30 IO/s: 270212 MB/s: 131 > nvmecontrol perftest -n 32 -o write -s 512 -t30 nvme0ns1 > Threads: 32 Size: 512 WRITE Time: 30 IO/s: 13658 MB/s: 6 I see datasheet specified performance metrics for 4K (not 512bytes) random r/w. Can you try this? This is may be impact write IO/s at least. > I was able to recover from the errors from previous message. They were > cause by the fact that i've commented hw.nvme.per_cpu_io_queues=0 in > loader.conf . After setting it back things got back to "slow normal" > Performance is half of what it should be and issue seems to be the nvme > driver. I'll try it on another OS to confirm it's not a hardware setting > issue.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20150116091317.GA83755>