e4rat - reduce boot time (into X) by some 50% for ext4

graysky · 2011-04-05 21:21:23

It does via slim/ldxe. Nothing too heavy. Trying it on an old workstation now.

toad · 2011-04-05 21:23:22

I'll whack up a couple of bootcharts when I get the time and start a thread, prolly in a couple of days.

graysky · 2011-04-05 21:30:34

Damn - no change on that older Athlon system. 21 sec with or without

toad · 2011-04-05 21:33:33

Again, how many files in your startup.log? Speed of machine is irrelevant, it is all about optimising disk access time and CPU load.

graysky · 2011-04-05 21:48:35

Nearly 1,100:

$ wc -l /var/lib/e4rat/startup.log
1075 /var/lib/e4rat/startup.log

toad · 2011-04-05 22:34:55

Okay, so I went a little over board with autologin KDE, FF4, Thunderbird, Skype, Kopete, Yakuake, Dropbox and wicd but all that took 2 mins 46 seconds as per normal and just under 90 seconds with e4rat.

My startup.log has no less than 3471 files in it and the machine I am using is a Thinkpad T61 with an Intel(R) Core(TM)2 Duo CPU T7500 @ 2.20GHz CPU with 2GB RAM.

Just under 50% reduction in start-up time is fine with me although the above is really something for a desktop, not a laptop.

Bootchart does not seem to do the job as it stops at kdm, but it is too late for me to look into it now.

hauzer · 2011-04-05 23:29:15

Pardon me if it's a noob question, but why/is this only for ext4?

falconindy · 2011-04-06 00:07:46

hauzer wrote:

Pardon me if it's a noob question, but why/is this only for ext4?

From the first post:

e4rat reduces disk access times through physical file reallocation. It is based
on the online defragmentation ioctl EXT4_IOC_MOVE_EXT from the ext4 filesystem,
which was introduced in Linux Kernel 2.6.31. Therefore, other filesystem types
or earlier versions of extended filesystems are not supported.

ChoK · 2011-04-06 00:11:32

The binaries should be in /sbin, not in /usr/sbin some people (like me) may have /usr in a different partition not mounted at startup. Similarly, If the .so are needed for the collect and preload binary they should be in /lib instead of /usr/lib.

PS : maybe devtmpfs will make this work mmm time to test.
edit : silly me, there is no way the FS will be mounted before fsck .... devtmpfs just populate the /dev directory before udev is up

Last edited by ChoK (2011-04-06 00:22:17)

TigTex · 2011-04-06 01:27:57

Tested this on my old mediacenter (pentium4 3.0ghz HT, 40gb ide 7200rpm, 1gb ddr) with xbmc... from grub to console, there was no improvement (8.4 seconds), but from console to xbmc, the diference is huge! X and xbmc start in +/- 2 seconds, before I had to wait at least 15 seconds! This sure works with me, thank you!
(I had 660 files on startup.log, minimal install + xbmc)

When I upgrade my kernel and do some other major changes, do I have to re-run e4rat-collect and alloc manually?
It would be nice to see this program to be used to improve most used programs startup time, just by reallocating the files in correct order (just like some disk optimizers/defrag do on windows).

ps: readme.install gives an error on line 22 because of a extra "}"

Last edited by TigTex (2011-04-06 01:29:59)

ngoonee · 2011-04-06 01:57:27

I wonder, systemd requires an init line and so does e4rat... how would I use both?

hauzer · 2011-04-06 02:02:19

falconindy wrote:

From the first post:

e4rat reduces disk access times through physical file reallocation. It is based
on the online defragmentation ioctl EXT4_IOC_MOVE_EXT from the ext4 filesystem,
which was introduced in Linux Kernel 2.6.31. Therefore, other filesystem types
or earlier versions of extended filesystems are not supported.

Thanks for that, my eyes are kinda heavy right now. If I would 'convert' ext3 to ext4 could it be possible for this to work?

fabertawe · 2011-04-06 08:02:15

2239 items... burg to login prompt - no change, 20 seconds. Login to working Gnome desktop has gone from 30 down to 15 seconds! Very impressed.

I also included Thunderbird and Firefox as I usually run them after logging in and they also start faster.

SanskritFritz · 2011-04-06 08:52:42

hauzer wrote:

Thanks for that, my eyes are kinda heavy right now. If I would 'convert' ext3 to ext4 could it be possible for this to work?

Yes, if you really convert your ext3 to ext4, and not just mount is as ext4.

ngoonee · 2011-04-06 12:10:08

SanskritFritz wrote:

hauzer wrote:
Thanks for that, my eyes are kinda heavy right now. If I would 'convert' ext3 to ext4 could it be possible for this to work?
Yes, if you really convert your ext3 to ext4, and not just mount is as ext4.

I'd like to hear from others how this conversion goes. My last non-ext4 partition is pretty important (all my data/work) and I don't really feel like restoring backups for a couple of hours.

falconindy · 2011-04-06 12:36:01

There's no shortage of information available on converting ext3 to ext4 just by enabling the extra features via tune2fs. However, it doesn't, for example, magically convert all your written data to use extents. I'm dubious that you'd get any immediate mileage out of e4rat by performing such a conversion.

Spider.007 · 2011-04-06 12:44:05

Has anyone compared the stats with preload?

ngoonee · 2011-04-06 13:03:52

falconindy wrote:

There's no shortage of information available on converting ext3 to ext4 just by enabling the extra features via tune2fs. However, it doesn't, for example, magically convert all your written data to use extents. I'm dubious that you'd get any immediate mileage out of e4rat by performing such a conversion.

That's what I thought. It wouldn't 'really' be an ext4 system that way, just an ext3 system which can do a few of the things ext4 can, is that right?

hauzer · 2011-04-06 13:21:57

Yes, I f*cked up my system testing this. The 'conversion' goes fine, but e4rat still can't move the files and they somehow get corrupted. Now I have corrupt libc.so and whatnot.. Oh well.

toad · 2011-04-06 16:26:07

hauzer wrote:

Yes, I f*cked up my system testing this. The 'conversion' goes fine, but e4rat still can't move the files and they somehow get corrupted. Now I have corrupt libc.so and whatnot.. Oh well.

That is bad! I've tested it with quite a few different start scenarios and had no probs. What do the logs say?

andr3as · 2011-04-06 17:14:32

I just tried your package, it works flawlessly for me. Thank you very much!
System startup feels a little bit faster, but KDE seems to start a lot quicker. I haven't analyzed any bootcharts yet, might do so later.

By the way, if you use runlevels to start your DM and just append the e4rat init stuff to the grub line: If your DM doesn't start any more, put the runlevel to the end of the line. (At least I had to put the '5' to the end, or the system would boot to runlevel 3...)

lilsirecho · 2011-04-06 17:32:27

ext4 only?

lilsirecho · 2011-04-06 17:33:49

x86_64 compatible?

KlavKalashj · 2011-04-06 18:03:49

lilsirecho wrote:

x86_64 compatible?

Works fine here on x86_64. Using KDE, it was a noticeable difference, I didn't use bootchart but I could feel it. KDM and desktop loading was really fast.

fsckd · 2011-04-06 19:17:56

boot takes a few seconds longer
but processes load much faster

i suppose one would have to reallocate every now and then
as package updates would remove files already reallocated
edit: monthly cron job :-)

Last edited by fsckd (2011-04-06 19:29:19)

Arch Linux

#26 2011-04-05 21:21:23

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#27 2011-04-05 21:23:22

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#28 2011-04-05 21:30:34

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#29 2011-04-05 21:33:33

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#30 2011-04-05 21:48:35

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#31 2011-04-05 22:34:55

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#32 2011-04-05 23:29:15

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#33 2011-04-06 00:07:46

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#34 2011-04-06 00:11:32

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#35 2011-04-06 01:27:57

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#36 2011-04-06 01:57:27

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#37 2011-04-06 02:02:19

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#38 2011-04-06 08:02:15

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#39 2011-04-06 08:52:42

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#40 2011-04-06 12:10:08

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#41 2011-04-06 12:36:01

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#42 2011-04-06 12:44:05

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#43 2011-04-06 13:03:52

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#44 2011-04-06 13:21:57

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#45 2011-04-06 16:26:07

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#46 2011-04-06 17:14:32

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#47 2011-04-06 17:32:27

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#48 2011-04-06 17:33:49

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#49 2011-04-06 18:03:49

Re: e4rat - reduce boot time (into X) by some 50% for ext4

#50 2011-04-06 19:17:56

Re: e4rat - reduce boot time (into X) by some 50% for ext4

Board footer