SystemD would definitely help but Canonical would have to get over their Not-invented-here attitude with upstart
Phoronix: Ubuntu 11.10 Boot Performance: It's Slowing Down
When Ubuntu 11.04 was released, Phoronix provided Bootchart results for five different systems showing the boot performance from Ubuntu 8.04.4 LTS through Ubuntu 11.04. Since Ubuntu 10.04 LTS, the boot time has unfortunately increased.
http://www.phoronix.com/vr.php?view=16465
SystemD would definitely help but Canonical would have to get over their Not-invented-here attitude with upstart
The maximum disk throughput is misleading.
You'd think more = better, but that's not true. It's only better assuming that the additional I/O was necessary.
It's hard to tell because of image compresion, but it seems that major diffrence on bootchart is apt demon (aptd). Maybe fact that it is snapshot release where ton of packages gets updated metters?
Second thing would be, that first start after update that changes boot process(fe: installing bootchart, or new kernel version) is slower(it gathers information to speed up subsequents boots). Did you measure second boot without applying any updates/installing packages, or first one?
Last edited by Nait; 09-28-2011 at 01:19 AM.
Right, well I don't know what's going on in ur testing, but I can promise u that the boot time has about halved on my inspiron 1420 since I dist-upgraded from 11.04 tp 11.10. Im leaving this from my phone, so I can't give udetails at the moment, but replacing gdm has seriously improved myafter boot performance, and my evidence of this is not from throughput or any such metric, but literally the time it takes to boot up, and after all isn't that all that matters? I believe its a 2.20 ghz core 2 duo with 2 gigs ddr2 and a plain old 320 gig hdd
it probably wouldn't. systemd does not magically make things faster; there isn't much architecturally about systemd you could point to and say 'this makes is faster than upstart'. that's not why lennart wrote it, the advantages (as he perceives it) are in other areas.
and no current general purpose distro is very close to an optimal boot process out of the box at all, really, whether they use upstart or systemd or something else. all distros could optimize boot speed much better by tweaking their init process than by switching init daemon. But it's not actually really that important to any distro for anything but PR purposes at present, though there are interesting niche cases.
Small nit pick on the sandy bridge:
Maximum disk throughput increased
Either it decreased or the numbers are inverted.from 238MB/s to 147MB/s.
Back on topic,
The maximum disk throughput is an interesting metric, but doesn't mean much in its self. Also, a more fine-grained analysis of what has changed would be interesting, and probable more useful to the ubuntu devs.
David
After being completely pissed off with ubuntu's boot time I've installed Gentoo, just to check whether it's me or ubuntu - with parallel boot in openrc my gentoo fires up in 10-15s. Then I've turned on verbose upstart and disabled plymouth (just to see what happens behind the scene) and I saw services constantly going up and down during the boot process, reacting on some events, triggering other events - that is to say making complete mess during the boot.
Hence I'd say replacing upstart with whatever else can really have a big impact on boot time.
P.S. kernel bootstrap time (time lapse from grub till (pre)init) on both kernels is about 7sec, although ubuntu is with initrd and after kernel boots it passes control to boot-* scripts in initrd, which takes another 3-5s, and in gentoo I compiled monolithic kernel with all required modules (except optional) compiled-in - hence without initrd.
Hi,
You should probably redo your benchmarks with the very latest version of Ubuntu 11.10. Your bootcharts clearly show 3 bugs that we've already fixed, which severely affected boot performance
https://launchpad.net/bugs/854101
https://launchpad.net/bugs/850055
https://launchpad.net/bugs/856475