[mythtv-users] Can a diskless frontend WOL the backend?

Paul Bender pebender at san.rr.com
Sun Jun 4 02:09:50 UTC 2006


Phill Edwards wrote:

> I always thought this looked a really interesting approach to
> improving boot times -
> http://www-128.ibm.com/developerworks/linux/library/l-boot.html. I
> haven't got around to trying it yet so can't vouch for it.
> 
> There's also some inrteresting info at
> http://enterprise.linux.com/article.pl?sid=06/04/13/173227&tid=129
> 
> NB - all of tehse approaches look like pretty serious changes - caveat
> emptor etc. Also, I imagine one of the drawbacks is that unless
> they're part of a distro you're going to have to re-apply all the good
> work every time you upgrade the distro.
> 
> If you do have any success please share!

I contribute to a custom MythTV frontend Linux distribution called 
MiniMyth <http://linpvr.org>. As part of this, I implemented some of 
these parallel booting techniques in order to see whether or not it 
reduce the boot time. For this specific application, I found that it had 
little impact on boot time so I did not include them.

I could see how it might have significant impact on the boot time for a 
server that starts numerous services. However, a dedicated MythTV 
frontend Linux distribution starts very few services.

For MiniMyth booting from a USB key on an ASUS A8N-VM CSM motherboard, 
the entire boot time is 40s. However, the BOIS takes 5s, loading the 
kernel and image takes 10s, detecting and mounting the USB drive takes 
5s, and starting X and mythfrontend takes 5s. That leaves less than 15s 
of boot time that might be run in parallel. Therefore, there is very 
little time to be saved by making these things run in parallel.

For MiniMyth booting over the network, it takes somewhat longer (55s for 
the ASUS A8N-VM CSM motherboard) and running things in parallel saved 
slightly more time. However, the time saved was still not a very 
significant portion of the boot time.


More information about the mythtv-users mailing list