Looking at Hardened

anon31687413 · August 27, 2020, 1:26pm

Ok, well then you can post your systemd-analyze values
It’s not like I don’t believe you, I just want to find out why it boots faster because it doesn’t (yet) make sense to me.
In raw performance, it is most certainly not faster.

Beardedgeek72 · August 27, 2020, 1:28pm

Startup finished in 3.234s (kernel) + 6.084s (userspace) = 9.318s 
graphical.target reached after 5.927s in userspace

Nothing overly exciting. Tho I dual boot with W10, so there is a sliiiight difference there…
Also, we have at least one other person here (Judd?) who has the same experience vs the Zen kernel.

anon31687413 · August 27, 2020, 1:30pm

Thanks. I think I’ll give it a try as well, and report back.

pebcak · August 27, 2020, 1:37pm

I’ve got both zen and hardened installed. What kind of benchmark can I run to be able to compare the performance? Would you be kind to point out some to me?

linux-hardened 5.7.18.a-1
linux-zen 5.8.4.zen1-1 (from testing)

Resiliencia · August 27, 2020, 1:37pm

$ systemd-analyze && uname -rs
Startup finished in 6.207s (kernel) + 3.136s (userspace) = 9.344s 
graphical.target reached after 2.707s in userspace
Linux 5.7.18.a-1-hardened

anon31687413 · August 27, 2020, 1:37pm

You can use my benchmarker script:

github.com

torvic9/bash-scripts/blob/master/mini-benchmarker/benchmarker.sh

#!/bin/bash
# by torvic9
# Contributors and testers:
# Richard Gladman, William Pursell, SGS, mbb, mbod, Manjaro Forum, StackOverflow

runstress1() {
	local RESFILE="$WORKDIR/runstress1"
	/usr/bin/time -f %e -o $RESFILE $STRESS -q $WORKDIR/stressC &>/dev/null &
	local PID=$!
	echo -n -e "* stress-ng cpu:\t\t\t"
	local s='-+'; local i=0; while kill -0 $PID &>/dev/null ; do i=$(( (i+1) %2 )); printf "\b${s:$i:1}"; sleep 1; done
	printf "\b " ; cat $RESFILE
	echo "stress-ng cpu: $(cat $RESFILE)" >> $LOGFILE
	return 0
}

runstress2() {
	local RESFILE="$WORKDIR/runstress2"
	/usr/bin/time -f %e -o $RESFILE $STRESS -q $WORKDIR/stressR &>/dev/null &
	local PID=$!

This file has been truncated. show original

anon31687413 · August 27, 2020, 1:39pm

Here’s the boot time comparison:

linux:
Startup finished in 3.098s (kernel) + 6.863s (userspace) = 9.962s

linux-hardened:
Startup finished in 3.123s (kernel) + 6.600s (userspace) = 9.723s

linux-vd:
Startup finished in 2.967s (kernel) + 4.901s (userspace) = 7.869s

So, hardened indeed boots 0.2s faster, but this could be down due to the fact that it’s a 5.7 kernel.
Also shows that custom kernels can really improve boot time

EDIT: I don’t believe the numbers for linux-vd. Going to analyze this.

Hystrix · August 27, 2020, 1:42pm

What about zen ??

pebcak · August 27, 2020, 1:43pm

Shall I run it as root?

Resiliencia · August 27, 2020, 1:45pm

It’s good, but it’s also essential to clear the cache and especially the journal time
du /var/log/journal -h
as well as check sudo paccache -d then the starting times are faster

anon31687413 · August 27, 2020, 1:45pm

No, never. Make sure you’ve installed the required dependencies:

depends=('make' 'cmake' 'time' 'perf' 'unzip'
	 'nasm' 'inxi' 'argon2' 'gmp' 'wget' 'blender')

anon31687413 · August 27, 2020, 1:46pm

Which cache?
Package cache? In which way does that have an influence on boot time?

Resiliencia · August 27, 2020, 1:50pm

 $ sudo paccache -d
==> no candidate packages found for pruning

$ du /var/log/journal -h
17M     /var/log/journal/a6e8f7216afb42bd93573d9f10d50a19
17M     /var/log/journal

anon31687413 · August 27, 2020, 1:51pm

ldconfig.service, the dynamic linker cache, sometimes gets rebuilt on boot, it has a large influence on boot time.
So you might all want to check systemctl status ldconfig.service and see if it ran on boot.

anon31687413 · August 27, 2020, 1:52pm

You don’t need sudo for paccache -d.

To check size of the journal, use journalctl --disk-usage.

Resiliencia · August 27, 2020, 1:52pm

Believe it or not, sometimes when I don’t clear the packet cache, the time rises to 17-20 or more seconds at startup.

Hystrix · August 27, 2020, 1:53pm

Off topic

I have reinstalled EOS on Btrfs and it feels faster . Boot time is decreased
and firefox starts in half the time .

anon31687413 · August 27, 2020, 1:53pm

Package cache sits on a btrfs subvolume here.

Resiliencia · August 27, 2020, 1:57pm

$ journalctl --disk-usage Archived and active journals take up 16.0M in the file system.
It is also valid your form, of course

anon31687413 · August 27, 2020, 2:23pm

Do you have compression enabled? Maybe this could explain faster startup.