Further questions on performance of IBM440GX/GP regarding boot time

Further questions on performance of IBM440GX/GP regarding boot time

[Markus Bernauer]

Hi Eugene, hello NG,

I was following the discussion about the performance of  different
IBM440GX/GP systems with increasing interest.
We are going to use an IBM400GP in conjunction with the Ebony-Evalboard
for a project that demands boot times as short as possible. Could you
please give me any hints about how fast your systems boot and whether
you optimized the systems to reach these bootup times.
I know that there are several possibilites to reduce boot time of linux,
but at the moment it would suffice to have some values for orientation.
I also know that there are differences in boot time dependent on
hardware and software configuration so it would be nice to get
at least little information about the systems the different boot times
belong to.
The other, less preferable possibility to achieve fast bootup, would be
to use SCIOPTA. But I am afraid this would be off topic, isn´t it?;-)

Thanks in advance

Markus


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Further questions on performance of IBM440GX/GP regarding boot time

[Eugene Surovegin]

On Thu, Mar 25, 2004 at 12:44:30PM +0100, Markus Bernauer wrote:
> I was following the discussion about the performance of  different
> IBM440GX/GP systems with increasing interest.
> We are going to use an IBM400GP in conjunction with the Ebony-Evalboard
> for a project that demands boot times as short as possible. Could you
> please give me any hints about how fast your systems boot and whether
> you optimized the systems to reach these bootup times.

Well, I cannot say I spent any time optimizing boot times :).

My estimation is 1-2s for bootloader, 2-3s for Linux kernel, provided you kernel
is in the flash memory, and doesn't have a lot of hw to detect (like SCSI
disks). Also note that time I gave to the kernel doesn't include startup
scripts (everything started by init). Usually init processing takes a lot of
time, so using something like busybox helps a lot.

One thing, to achieve such boot speeds _disable_ serial system console.

My estimations are based on 405GPr-based hw boot times.

> I know that there are several possibilites to reduce boot time of linux,
> but at the moment it would suffice to have some values for orientation.
> I also know that there are differences in boot time dependent on
> hardware and software configuration so it would be nice to get
> at least little information about the systems the different boot times
> belong to.
> The other, less preferable possibility to achieve fast bootup, would be
> to use SCIOPTA. But I am afraid this would be off topic, isn?t it?;-)

Hmm, I have no idea what SCIOPTA is :).

Eugene

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/