init observations

Jeff Spaleta jspaleta at gmail.com
Wed Nov 16 00:04:15 UTC 2005


On 11/15/05, Dave Jones <davej at redhat.com> wrote:
> If suspend was 100% reliable, I'd agree.  For a lot of systems,
> we may never get them working due to lack of hardware specifications
> to wake the drivers up on resume. (Video drivers are particularly
> bad in this area).

I'm sure its going to be the hard thing to do... but who said the
right thing to do was the easy one?  I'm sure the difficulties in
hardware coverage are real, but I still haven't seen a smoking gun
that init parallelization is going to be an instant win for boottime
for the default case, the average case, the worst case or anything
else worth noting.  The introduction of the bootchart and the
resulting optimization of the init process we have now and I don't see
why parallelization is garunteed to improve on that. Considering all
the other technical benefits to moving to a more sophisticated init, I
just don't think boottime is a strong argument.. I don't think we
really know what the impact on boottime is going to be.

-jef"I thought the mark was 80% reliable"spaleta




More information about the devel mailing list