Malinux' notes: Difference between revisions
Line 67: | Line 67: | ||
</syntaxhighlight> | </syntaxhighlight> | ||
You know have changed scheduler to bfq | You know have changed scheduler to bfq | ||
To make this survive a reboot you can add this snippet from the arch wiki that | To make this survive a reboot you can add this snippet I borrowed from the arch wiki that | ||
set scheduler based on hard drive type. | set scheduler based on hard drive type. none for nvme, mq-deadline for ssd and none for nvme | ||
Add the snippet to /etc/udev/rules.d/60-ioschedulers.rules | |||
<syntaxhighlight lang="bash" | <syntaxhighlight lang="bash" | ||
# set scheduler for NVMe | # set scheduler for NVMe | ||
Line 76: | Line 77: | ||
# set scheduler for rotating disks | # set scheduler for rotating disks | ||
ACTION=="add|change", KERNEL=="sd[a-z]*", ATTR{queue/rotational}=="1", ATTR{queue/scheduler}="bfq" | ACTION=="add|change", KERNEL=="sd[a-z]*", ATTR{queue/rotational}=="1", ATTR{queue/scheduler}="bfq" | ||
</syntaxhighlight | </syntaxhighlight> | ||
source: https://wiki.archlinux.org/title/Improving_performance#Changing_I/O_scheduler | source: https://wiki.archlinux.org/title/Improving_performance#Changing_I/O_scheduler |
Revision as of 14:05, 21 November 2021
HDD hacks
Disable NCQ
Disabling NCQ (Native Command Queuing) can in some circumstances improve performance on HDD's
# To check if ncq is already off, its' likely not if you haven't already turned it off
cat /sys/block/sdX/device/queue_depth
# it will return 32 if it's on and 1 if it's turned off
# To turn it off:
echo 1 > /sys/block/sdX/device/queue_depth
To have this survive a reboot and disable ncq for all your hdd's or ssd's we can make a systemd script First we'll make a bash script that the systemd script run
#!/bin/bash
disker=$( lsblk -d | awk '/^sd/ { print $1 }' )
# echo $disker
for disk in $disker
do
q="/sys/block/$disk/device/queue_depth";
newq=1;
echo $newq > $q;
done
Put the script in /usr/local/bin/disable_ncq.sh and make it executable
chmod +x /usr/local/bin/disable_ncq.sh
Then we go on to make the systemd script itself.
Create /etc/systemd/system/disable-ncq.service and add:
# vim:isfname-==
[Unit]
Description=HDD queuing
[Service]
Type=oneshot
ExecStart=/usr/local/bin/disable_ncq.sh
[Install]
WantedBy=multi-user.target
Then we need to enable and start the script:
systemctl enable disable-ncq.sh
systemctl start disable-ncq.sh
BFQ I/O scheduler
I use BFQ btw. BFQ is a bloody fast I/O scheduler, best suitable with HDD's.
/sys/block/sdX/queue/scheduler
# the output from the cat command now will show
mq-deadline [bfq] none
You know have changed scheduler to bfq To make this survive a reboot you can add this snippet I borrowed from the arch wiki that set scheduler based on hard drive type. none for nvme, mq-deadline for ssd and none for nvme Add the snippet to /etc/udev/rules.d/60-ioschedulers.rules <syntaxhighlight lang="bash"
- set scheduler for NVMe
ACTION=="add|change", KERNEL=="nvme[0-9]n[0-9]", ATTR{queue/scheduler}="none"
- set scheduler for SSD and eMMC
ACTION=="add|change", KERNEL=="sd[a-z]*|mmcblk[0-9]*", ATTR{queue/rotational}=="0", ATTR{queue/scheduler}="mq-deadline"
- set scheduler for rotating disks
ACTION=="add|change", KERNEL=="sd[a-z]*", ATTR{queue/rotational}=="1", ATTR{queue/scheduler}="bfq" </syntaxhighlight>
source: https://wiki.archlinux.org/title/Improving_performance#Changing_I/O_scheduler