On 03/20/2016 14:30, Richard W.M. Jones wrote:
Thanks, this is much better than looking at 'ts -i' output.
BTW I have use 'git am' to apply them to latest git HEAD and try it.
Is this the preferred way, or is there a git branch/repo somewhere that I missed which
would already have these applied?
Not running the 'hwclock' command reduces boot times considerably.
However I'm not sure if it is safe. See the question I posted on
qemu-devel:
http://thread.gmane.org/gmane.comp.emulators.qemu/402194
At the moment, about 50% of the time is consumed by SeaBIOS. Of this,
about ⅓rd is SGABIOS (it contains a useless ¼s sleep waiting for a
keypress ... ffs!)
Ah, I didn't have sgabios installed, and it complained it couldn't find an event
(qemu:overhead?).
Installed it and I got similar results to yours, the largest overhead is BIOS:
https://gist.github.com/anonymous/37cfb3d4eb3d3a1c86b2
Thought to try booting a Xen PV domain for comparison, but AFAICT libguestfs doesn't
support LIBGUESTFS_BACKEND=libvirt:xen:///
, and the other ⅔rds is something else in SeaBIOS.
Simply removing SGABIOS improves boot times to below 2s, but at a cost
that we cannot see any messages from SeaBIOS so further measurement
and therefore improvement becomes impossible. I'm going to try to fix
SeaBIOS/SGABIOS first. See also:
http://thread.gmane.org/gmane.comp.emulators.qemu/402196
Nice, this would benefit booting regular VMs too, not just libguestfs appliances, right?
Best regards,
--
Edwin Török | Co-founder and Lead Developer
Skylable open-source object storage: reliable, fast, secure
http://www.skylable.com