Nixos benchmark?

Any one did benchmark on nixos comparing its performance with other distros (like ubuntu, fedora or arch linux). As I searched the internet but did find nothing.

2 Likes

NixOS should potentially be a little bit slower because of the indirection with huge paths and symlinks, other than that all distros are the same. It’s just about what packages you’re installing and using.

What would be more important is to tune your system accordingly. With Nix this might be easier than with other systems as your configuration remembers all your parameters.

2 Likes

mimicking the ‘speed hacks’ that clear linux may be achievable, with recompilation of packages with specific compiler optimisations and custom kernel/userspace patches.

https://clearlinux.org/about

I was going to give it a whirl, but i thought if i’m going to learn a lot of new paradigms about a ecosystem, i might as well go the nix route. I think i made the right decision.

optimisation is sometimes about tinkering, changing , patching , recompiling and updating things… something nix can do without fear of complete system collapse.

it was discussed in depth here…

I think needs extensive engineering effort , something were not exactly in abundance of.

3 Likes

There are some odd defaults going on with Linux kernel flags, started looking into it but I kinda forgot about it GitHub - gytis-ivaskevicius/linux-kernel-comparison It might be impacting performance in some areas.

Also, there is something going on with systemd, the difference is noticeable on low-end hardware boot time, especially if you were to look into NetworkManager/networkd startup time. It is not even comparable to Arch

1 Like

Here are a couple of screenshots from when I started investigating it a little (~06/04/2021). Not sure about either system configuration at this point.

NixOS:


Arch:


Not much of a reference point of anything, but that’s the best I got on hand

2 Likes

Don’t know how relevant the information I’ll be sharing here is, but here is more information for my previous setup (the images Gytis linked) + all the benchmarks we conducted:

Scheduler:

❯ cat /sys/block/sda/queue/scheduler
mq-deadline kyber [bfq] none

Both NixOS & Arch scaling_governor:

λ cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor
schedutil
schedutil
schedutil
schedutil
schedutil
schedutil
schedutil
schedutil

Drishal input on the subject:

Arch:
#
# CPU Frequency scaling
#
CONFIG_CPU_FREQ=y
CONFIG_CPU_FREQ_GOV_ATTR_SET=y
CONFIG_CPU_FREQ_GOV_COMMON=y
CONFIG_CPU_FREQ_STAT=y
# CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE is not set
# CONFIG_CPU_FREQ_DEFAULT_GOV_POWERSAVE is not set
# CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE is not set
CONFIG_CPU_FREQ_DEFAULT_GOV_SCHEDUTIL=y
CONFIG_CPU_FREQ_GOV_PERFORMANCE=y
CONFIG_CPU_FREQ_GOV_POWERSAVE=y
CONFIG_CPU_FREQ_GOV_USERSPACE=y
CONFIG_CPU_FREQ_GOV_ONDEMAND=y
CONFIG_CPU_FREQ_GOV_CONSERVATIVE=y
CONFIG_CPU_FREQ_GOV_SCHEDUTIL=y

NixOS:
#
# CPU Frequency scaling
#
CONFIG_CPU_FREQ=y
CONFIG_CPU_FREQ_GOV_ATTR_SET=y
CONFIG_CPU_FREQ_GOV_COMMON=y
# CONFIG_CPU_FREQ_STAT is not set
CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y
# CONFIG_CPU_FREQ_DEFAULT_GOV_POWERSAVE is not set
# CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE is not set
# CONFIG_CPU_FREQ_DEFAULT_GOV_SCHEDUTIL is not set
CONFIG_CPU_FREQ_GOV_PERFORMANCE=y
CONFIG_CPU_FREQ_GOV_POWERSAVE=m
CONFIG_CPU_FREQ_GOV_USERSPACE=m
CONFIG_CPU_FREQ_GOV_ONDEMAND=m
CONFIG_CPU_FREQ_GOV_CONSERVATIVE=m
CONFIG_CPU_FREQ_GOV_SCHEDUTIL=y

A while later Gytis submitted the following diff between the two kernel configs:

And this is what my systemd-analyze blame looks like today:

25.364s systemd-journal-flush.service
23.819s home-manager-icy\x2dthought.service
 9.066s bluetooth.service
 8.476s systemd-logind.service
 7.713s NetworkManager.service
 7.703s iwd.service
 6.145s firewall.service
 4.091s systemd-suspend.service
 3.986s polkit.service
 3.954s dev-disk-by\x2dlabel-nixos.device
 3.008s systemd-modules-load.service
 2.964s systemd-udevd.service
 2.727s mount-pstore.service
 2.683s systemd-fsck@dev-disk-by\x2dlabel-boot.service
 2.605s home.mount
 2.408s upower.service
 1.345s resolvconf.service
 1.328s modprobe@drm.service
 1.207s cpufreq.service
 1.125s boot.mount
 1.095s udisks2.service
 1.094s systemd-remount-fs.service
  922ms sys-kernel-debug.mount
  883ms systemd-backlight@backlight:amdgpu_bl0.service
  867ms dev-mqueue.mount
  864ms dev-hugepages.mount
  672ms sys-kernel-config.mount
  571ms systemd-random-seed.service
  557ms systemd-tmpfiles-setup-dev.service
  498ms sys-fs-fuse-connections.mount
  480ms audit.service
  459ms systemd-journald.service
  449ms dev-disk-by\x2dlabel-swap.swap
  387ms user@1000.service
  357ms modprobe@fuse.service
  344ms systemd-sysctl.service
  331ms systemd-tmpfiles-setup.service
  238ms systemd-tmpfiles-clean.service
  331ms systemd-tmpfiles-setup.service
  238ms systemd-tmpfiles-clean.service
  222ms modprobe@configfs.service
  206ms autorandr.service
  149ms network-setup.service
  149ms systemd-udev-trigger.service
  121ms systemd-timesyncd.service
  107ms network-local-commands.service
   81ms systemd-rfkill.service
   70ms systemd-update-utmp.service
   68ms rtkit-daemon.service
   50ms display-manager.service
   28ms systemd-user-sessions.service
   22ms nscd.service
   21ms post-resume.service
   19ms user-runtime-dir@1000.service
   10ms kmod-static-nodes.service
   10ms NetworkManager-dispatcher.service
    9ms pre-sleep.service
    2ms tmp.mount
2 Likes

what ever you do , don’t turn into a gentoo ricer. :-), i admit i might of been one when i was young and foolish. (now i’m just old and foolish).

https://www.shlomifish.org/humour/by-others/funroll-loops/Gentoo-is-Rice.html

Hosted by Flying Circus.