Martin Pitt [2010-10-04  9:05 +0200]:
> Indeed it seems that maverick dramatically regressed here, at least on
> my end. In August I got a nice 8 second boot [1], which already wasn't
> optimal (ureadahead shouldn't have blocked the boot), I
> now have a 30 second boot [2]
> [1] 
> http://people.canonical.com/~pitti/bootcharts/donald-maverick-20100805-1.png

Quick followup to unpanic: I did a fresh reinstall of my laptop a few
days ago, which apparently got rid of some cruft and the changed swap
partition, and I'm back at 12.9 s:

  http://people.canonical.com/~pitti/bootcharts/donald-maverick-final.png

Still worse than [1], though. It however shows nicely where there is
still optimization potential, since most of the time this is neither
CPU nor IO bound. It also shows some new things which crept into the
boot process:

  * The dreaded loadkeys/sh processes is still there, and worse, it's
    there twice 

  * There is a run-parts block which runs lsb_release,
    update-motd-cpu, check-bios-nx, awk, apt-config, and
    update-motd-update. Surely we don't need all this during boot?

  * tclsh?? I need to find out where that comes from; this will hit
    lower-power systems a lot

  * usb-modeswitch was something we introduced in maverick, but it's
    not supposed to run on systems without an USB 3G modem. (I do have
    an internal one, but it doesn't need to be switched)

  * (note) the desktop part of the chart is totally useless, it
    misassigns names to processes. I suppose the big
    gdm-session-worker process is actually compiz, I don't see that
    anywhere else.

This should be some nice fodder for further optimization work.

Martin
-- 
Martin Pitt                        | http://www.piware.de
Ubuntu Developer (www.ubuntu.com)  | Debian Developer  (www.debian.org)

Attachment: signature.asc
Description: Digital signature

-- 
ubuntu-devel mailing list
[email protected]
Modify settings or unsubscribe at: 
https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel

Reply via email to