Re: [gentoo-soc] (draft) final report for OpenRC soc project 2012

[Rich Freeman]

On Fri, Aug 17, 2012 at 3:29 AM, Luca Barbato <lu_zero@g.o> wrote:
> On 8/17/12 7:39 AM, heroxbd@×××××.com wrote:
>> Thanks a lot for the comment. I agree. Nobody cares about the tiny (not
>> tested) booting up time gain with rc_parallel.
>
> It can be not so tiny, surely busybox+openrc gives a better gain in many
> cases.
>

I suspect that it will depend greatly on what services you're running,
and what order they happen to start in, and what you care about.  In
theory slamming the kernel with a ton of processes will allow it to
manage its queues better with a fuller understanding of demand.
systemd can potentially short-cut this a bit further since it can
consider a dependency resolved if nothing more than a socket is
created, which is a clever trick (I have no idea how well it works out
in practice, though I have used a .socket service once and that worked
out fine (with the caveat that the first connection fails)).

Where I saw the bigger performance difference between openrc and
systemd was in shutdown.  Systemd shuts down REALLY fast, and I've
noticed it tends to actually kill ssh sessions rather than leaving
them unresponsive (plus ssh dies even faster so that makes it seem
subjectively faster - but the difference in full shutdown is still
real).  Obviously people don't care as much about shutdown
performance, but on a laptop (where supposedly event-driven inits
should shine) I know that waiting for a full shutdown before packing
up can be painful at least in the land of Windows.

Just food for thought - it seems odd that parallel startup should have
any issues at all, provided that service scripts don't terminate until
the service is functional and all dependencies are specified.  If
either of those aren't true then you'll have race conditions.

Rich