| View previous topic :: View next topic |
| Author |
Message |
Naib
Watchman
Joined: 21 May 2004
Posts: 6073
Location: Removed by Neddy
|
 Posted: Tue Oct 29, 2019 12:16 pm Post subject: The Irony of systemd - boottime |
 |
|
We Need to Talk About Systemd
Boot Time Optimization for the new init daemon
https://static.sched.com/hosted_files/osseu19/58/systemd-csimmonds-elce-2019.pdf
remember when Systemd was sold as a fast init system (because in-car needs fast)
_________________
#define HelloWorld int
#define Int main()
#define Return printf
#define Print return
#include <stdio>
HelloWorld Int {
Return("Hello, world!\n");
Print 0; |
|
| Back to top |
|
 |
mike155
Advocate
Joined: 17 Sep 2010
Posts: 4438
Location: Frankfurt, Germany
|
| Posted: Tue Oct 29, 2019 12:33 pm Post subject: |
|
|
That's a nice presentation. It explains what Systemd is and it shows how you can speed up boot time using Systemd tools.
There's nothing wrong with that? |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 55112
Location: 56N 3W
|
| Posted: Tue Oct 29, 2019 12:53 pm Post subject: |
|
|
mike155,
It goes even faster if you throw away systemd and friends.
-- edit --
Does systemd have any friends?
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
Anon-E-moose
Watchman
Joined: 23 May 2008
Posts: 6258
Location: Dallas area
|
| Posted: Tue Oct 29, 2019 12:58 pm Post subject: |
|
|
I never had a problem with boot times even before systemd showed up. *shrugs*
_________________
UM780, 6.12 zen kernel, gcc 13, openrc, wayland |
|
| Back to top |
|
|
mike155
Advocate
Joined: 17 Sep 2010
Posts: 4438
Location: Frankfurt, Germany
|
| Posted: Tue Oct 29, 2019 1:07 pm Post subject: |
|
|
| NeddySeagoon wrote: | mike155,
It goes even faster if you throw away systemd and friends.
|
That's true! And that's exactly what the presentation says (page 7, slide 17):
- Boot-up speed of Buybox init: Fast
- Boot-up speed of System V init: Slow
- Boot-up speed of Systemd: Medium
It also says that Systemd is 'not just an init system', but 'a way life' (page 8, slide 18 )
I can't find anything wrong in this presentation. Why don't you like it?
Last edited by mike155 on Tue Oct 29, 2019 6:58 pm; edited 1 time in total |
|
| Back to top |
|
|
Jaglover
Watchman
Joined: 29 May 2005
Posts: 8291
Location: Saint Amant, Acadiana
|
| Posted: Tue Oct 29, 2019 1:09 pm Post subject: |
|
|
When I boot my desktop most of the time takes firmware initializing hardware. Will systemd convince the firmware to boot faster? Once the boot is handed over to kernel everything is blistering fast with OpenRC, thanks to eMMC methinks.
_________________
My Gentoo installation notes.
Please learn how to denote units correctly! |
|
| Back to top |
|
|
erm67
l33t
Joined: 01 Nov 2005
Posts: 653
Location: EU
|
| Posted: Tue Oct 29, 2019 1:56 pm Post subject: |
|
|
An old classic:
https://www.youtube.com/watch?v=4NXMmHYNYfA
_________________
Ok boomer
True ignorance is not the absence of knowledge, but the refusal to acquire it.
Ab esse ad posse valet, a posse ad esse non valet consequentia
My fediverse account: @erm67@erm67.dynu.net |
|
| Back to top |
|
|
Ant P.
Watchman
Joined: 18 Apr 2009
Posts: 6920
|
| Posted: Tue Oct 29, 2019 6:23 pm Post subject: |
|
|
| The fact that the author thinks System V init is "slow", or that it "requires bash", really highlights how absolutely ignorant they are about how it works and where it begins and ends. Claiming busybox is zero files and zero MB too. They're drunk on the redhat kool-aid. |
|
| Back to top |
|
|
pjp
Administrator
Joined: 16 Apr 2002
Posts: 20595
|
| Posted: Tue Oct 29, 2019 7:56 pm Post subject: |
|
|
| NeddySeagoon wrote: | | Does systemd have any friends? |
Apparently most installations of Linux systems and at least one BSD developer. Anyone else fears change. The Tragedy of systemd
_________________
Quis separabit? Quo animo? |
|
| Back to top |
|
|
Anon-E-moose
Watchman
Joined: 23 May 2008
Posts: 6258
Location: Dallas area
|
| Posted: Tue Oct 29, 2019 8:57 pm Post subject: |
|
|
Interesting, I hadn't seen that before, and he has some valid points.
_________________
UM780, 6.12 zen kernel, gcc 13, openrc, wayland |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 55112
Location: 56N 3W
|
| Posted: Tue Oct 29, 2019 9:23 pm Post subject: |
|
|
pjp,
Thank you for that link. There are a lot of valid points there.
Being an old fart and systemd doesn't solve any problem I have, I don't want to invest my time in learning to do what I do today differently.
For me, its change for change sake.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
Anon-E-moose
Watchman
Joined: 23 May 2008
Posts: 6258
Location: Dallas area
|
| Posted: Tue Oct 29, 2019 10:03 pm Post subject: |
|
|
It's not a problem that all LP's ideas are ripped off from apple, at least to me, it's the fact that all the software that he "develops" is done in a half-assed way.
No engineering, no discipline, no real forethought (that I can tell), just a hacker mentality (gee whiz this is neat)
And like Neddy, I don't want to change just to change, unless there's benefits (lots of them) I don't want spend the effort learning something new.
And I hate monolithic software (you must install all of it) bah humbug.
_________________
UM780, 6.12 zen kernel, gcc 13, openrc, wayland |
|
| Back to top |
|
|
389292
Guru
Joined: 26 Mar 2019
Posts: 504
|
| Posted: Tue Oct 29, 2019 10:53 pm Post subject: |
|
|
openrc and systemd boot takes about the same on my machine.
however systemd's poweroffs are 3 minutes slower than openrc most of the times on my machine. |
|
| Back to top |
|
|
GDH-gentoo
Veteran
Joined: 20 Jul 2019
Posts: 1845
Location: South America
|
| Posted: Tue Oct 29, 2019 11:17 pm Post subject: |
|
|
| mike155 wrote: | - Boot-up speed of Buybox init: Fast
- Boot-up speed of System V init: Slow
- Boot-up speed of Systemd: Medium
[...]
I can't find anything wrong in this presentation. Why don't you like it? |
At the risk of being captain obvious, the irony here is that the presentation even exists. As you might know, at the time adoption of systemd was being pushed, one of the selling points that was used a lot was that it made the boot process faster. And here we have a presentation that not only says, in the author's opinion, that one init system that predates it is actually faster, but also that they felt the need to actively do something to try to reduce systemd's boot time. And, in that light, the fact that, after what appears to mostly consist on disabling useless units for the target embedded system, so that systemd does not waste time loading and activating them, the conclusion is that the boot process is still too long, is just hilarious.
Then, elaborating on what Ant P. said about the slide where systemd is compared to BusyBox init and "System V" init, to someone with the same level of knowledge about them as the one the author appears to have about systemd, that table just looks like a comparison of apples to oranges. The existence of a "required shell" row is outright wrong. Hint: Busybox and sysvinit, defined as the software packages that contain the corresponding "init daemon", don't need a shell at all. Let alone a particular implementation of it, like ash or bash. And saying that Busybox has 0 executables is... ugh. One wonders how the "init daemon" is implemented then, if it is not with an executable. Magic dust?
However, I do realize that the comparison is not the presentation's main point. I would have preferred a slide that just said someting like "we found that systemd was not fast enough, and that distribution XXX, that uses BusyBox init as process 1, actually booted faster, so here's what we did". And I agree that, as a brief introduction about systemd's init system aspects, it is fine. |
|
| Back to top |
|
|
Naib
Watchman
Joined: 21 May 2004
Posts: 6073
Location: Removed by Neddy
|
| Posted: Tue Oct 29, 2019 11:55 pm Post subject: |
|
|
And that's what made me laugh. I remember the PR when systemd was coming on the scene how it was faster than sysvinit and this was needed for incar entertainment.
Whether it was ignorance pointing to sysvinit (which is fast for what? 200 lines) or whether they didn't want to bring focus onto sysvrc and redhat appalling sh writing who knows BUT people bought into and well it is what it is.
It is just soo amusing also hearing that incar entertainment uses busybox for speed because systemd is soo slow when this was the example usecase to push systemd
In other news Debian are having a dev vote whether to support a non-systemd system as it is splitting into those that can't be bothered (sysd does it all for them) and those that are worried about the encapsulation
_________________
#define HelloWorld int
#define Int main()
#define Return printf
#define Print return
#include <stdio>
HelloWorld Int {
Return("Hello, world!\n");
Print 0; |
|
| Back to top |
|
|
pjp
Administrator
Joined: 16 Apr 2002
Posts: 20595
|
| Posted: Wed Oct 30, 2019 12:30 am Post subject: |
|
|
| Anon-E-moose wrote: | | Interesting, I hadn't seen that before, and he has some valid points. |
| NeddySeagoon wrote: | pjp,
Thank you for that link. There are a lot of valid points there.
Being an old fart and systemd doesn't solve any problem I have, I don't want to invest my time in learning to do what I do today differently.
For me, its change for change sake. |
You're welcome. I too thought most (maybe all) of the points were valid. I was disturbed by the superficial dismissal of criticism (all software has bugs, etc. which I thought went along with most of the valid points). It also seemed a bit too much on the side of a sales pitch and a "one of us" chant. Not to mention "talking down" to anyone that "fears change."
I haven't seen that it solves any problems that concern me, and I don't agree that what is good for RH is good for everyone / everything else. Otherwise I really don't care (I liked Solaris' SMF, but thought the use of xml was a pain).
On the other hand, I can apparently save ~3hrs a year if I shutdown and reboot on daily basis.
_________________
Quis separabit? Quo animo? |
|
| Back to top |
|
|
CaptainBlood
Advocate
Joined: 24 Jan 2010
Posts: 3998
|
|
| Back to top |
|
|
ct85711
Veteran
Joined: 27 Sep 2005
Posts: 1791
|
| Posted: Wed Oct 30, 2019 6:45 am Post subject: |
|
|
| I find it interesting that even though now; I still haven't seen anything that systemd brings as a benefit of any way for my systems. Even on one of my systems that is running systemd, it hasn't helped on anything. If anything; I am constantly fighting it over it thinking I want some service when I don't; and/or won't letting me restart the dang service. As far as start up speeds, it is slower than my older system that is running openrc with 2x as many services starting/installed. The logs side? Nope, it's more of a hassle with systemd than anything. I like to skim through my logs once per week to look for something abnormal (rarely there is any issues). For my openrc system, I can easily open the logs in vim, and as fast as vim can scroll through, I can easily skim through to find an abnormal sectiion (easy to see, when 98% tends to be the same set of lines echo'd throughout the log). For systemd, I have to first have it dump everything to the terminal; find the top of the logs; and skim through at the speed of the terminals scrolling rate. |
|
| Back to top |
|
|
technotorpedo
Apprentice
Joined: 10 Dec 2019
Posts: 151
|
| Posted: Wed Dec 11, 2019 5:46 am Post subject: |
|
|
LMFAO !!! Lol ... luv it. Yep, a picture speaks 1,000 words, that's why braindead folks only ever look at pictures which agree with their zombie-like views and perceptions eh ? 
Systemd as with any other thing can also be tweaked, can/should be anyway, regardless of init in my opinion. Profiling and reviewing an Os's boot is just low hanging fruit, very quick, very easy to pluck some decent gains depending, at least with SysV and Systemd + similar. If someone actually has to write initscripts etc obviously going to be a bit more involved. An ABSOLUTE MUST DO for any nixer using any type of out-of-box distro. When was using SysV, used software like bum and rcconf to help me find and evaluate such junk. Now with systemd, yep ... using systemd-analyze blame, + critical-chain too. Still look at what junk is automatically starting on a systemd OS with rcconf etc. One thing on my 2do-list is converting any old initscript chuff into systemd service units and what-not. Will mean some small amount of increased stability, some tiny amount of gain to boot time using systemd too. When the backwards compatibility chuff is gone and a system is overall running fully systemd native. I look forward to such a system myself. You are of course free to look forward or keep looking back into the past as ye heart desires, shrugs. Your OS's(installs), your hardware, your choices.
OP, you obviously have some serious issues with systemd and/or Lennart P, ya may want to seek some therapy ? Challenge it(systemd) or Lennart to a fight to the death, dunno ? Or perhaps spend some time with a naked woman (offline.) Lol ... what, couldn't help it. Plenty of others were thinking it, I just typed it out loud.  |
|
| Back to top |
|
|
bunder
Bodhisattva
Joined: 10 Apr 2004
Posts: 5956
|
| Posted: Wed Dec 11, 2019 12:23 pm Post subject: |
|
|
openrc could be faster if parallel start wasn't buggy... if i have 16c32t, i should be able to use all of them at startup, instead of starting each service one by one single threaded. /shrug
_________________
| Neddyseagoon wrote: | | The problem with leaving is that you can only do it once and it reduces your influence. |
banned from #gentoo since sept 2017 |
|
| Back to top |
|
|
Anon-E-moose
Watchman
Joined: 23 May 2008
Posts: 6258
Location: Dallas area
|
| Posted: Wed Dec 11, 2019 1:02 pm Post subject: |
|
|
| technotorpedo wrote: | | OP, you obviously have some serious issues with systemd and/or Lennart P, ya may want to seek some therapy |
Techno sure sounds a lot like the other systemd troll that likes to post on these types of threads.
_________________
UM780, 6.12 zen kernel, gcc 13, openrc, wayland |
|
| Back to top |
|
|
Anon-E-moose
Watchman
Joined: 23 May 2008
Posts: 6258
Location: Dallas area
|
| Posted: Wed Dec 11, 2019 1:03 pm Post subject: |
|
|
| bunder wrote: | | openrc could be faster if parallel start wasn't buggy... if i have 16c32t, i should be able to use all of them at startup, instead of starting each service one by one single threaded. /shrug |
Even with single threading I haven't noticed booting being that slow, though I only boot/reboot 5-10 times a year on average.
_________________
UM780, 6.12 zen kernel, gcc 13, openrc, wayland |
|
| Back to top |
|
|
Tony0945
Watchman
Joined: 25 Jul 2006
Posts: 5127
Location: Illinois, USA
|
| Posted: Wed Dec 11, 2019 3:08 pm Post subject: |
|
|
| Anon-E-moose wrote: | | Even with single threading I haven't noticed booting being that slow, though I only boot/reboot 5-10 times a year on average. |
Likewise. I have a 20 second wait configured into reFind in order to choose a different kernel or device. The longest wait i can't control is the BIOS init. Once the bios init screen goes away and the refind screen appears I hit ENTER, stuff scrolls on the screen in a blur and about 4 to 5 seconds later I'm looking at the xdm login prompt. Maybe 15 seconds total. If systemd took zero time, it would still take ten or so seconds until systemd started.
Now the Raspberry pi takes a while, but even the old k6-3 doesn't take that long. Logged in with ssh, I type "reboot", then when the connection goes down "ping k6", when the ping starts to answer, I hit control-c, "ssh root@k6", et cetera. No, it's not instantaneous and I really doubt that systemd would be either on a 450Mhz machine with 600Meg RAM. It's Windoze, even in a virtual Machine that takes forever to come up. |
|
| Back to top |
|
|
Ant P.
Watchman
Joined: 18 Apr 2009
Posts: 6920
|
| Posted: Wed Dec 11, 2019 9:04 pm Post subject: |
|
|
| technotorpedo wrote: | | OP, you obviously have some serious issues with systemd and/or Lennart P, ya may want to seek some therapy ? |
Let's see, you signed up two days ago and since then have done nothing but waltz into threads like these to throw shit at people and gloat about how you're proudly ignorant of Linux and that you're not a gentoo user, with a rancidly obnoxious loquaciousness that still manages to come off as illiterate lazy AOL-kid slang. And here's a clear cut insult and a necro-bump to boot.
You don't plan on being unbanned from this community for very long, do you? Looking to score brownie points for whatever troll forum sent you here? |
|
| Back to top |
|
|
Naib
Watchman
Joined: 21 May 2004
Posts: 6073
Location: Removed by Neddy
|
| Posted: Wed Dec 11, 2019 10:29 pm Post subject: |
|
|
hey technotorpedo, do you want to know why a number of us are concern with the philosophy of systemd? it is because of things like this:
https://gitlab.gnome.org/GNOME/gnome-shell/commit/9a45d9692aba6d0ff7d30b0f73237b78b3281b68
| Quote: | | * Place launched applications into a systemd scope [Benjamin; !863] |
New version of GNOME (again) dependant upon an initsystem. Sure this might be a build-time OPTIONAL extra but this could end up being like logind... why does a desktop application depend on an init/service manager. If systemd kept to itself and offered itself as the best alternative, if it provided a truly modular system you would see people choosing it because it is the better option rather than enforced.
to then question the concept brought up in the 1st post clearly shows a COMPLETE lack of appreciation of the koolaid that systemd was sold as.
Troll account is a troll
_________________
#define HelloWorld int
#define Int main()
#define Return printf
#define Print return
#include <stdio>
HelloWorld Int {
Return("Hello, world!\n");
Print 0; |
|
| Back to top |
|
|
|
Gentoo Chat
