Faster is nice but ... [LWN.net]

Faster is nice but ...

Posted Sep 20, 2003 8:00 UTC (Sat) by ressu (subscriber, #14615)
In reply to: Faster is nice but ... by AnswerGuy
Parent article: Boot Linux faster (IBM developerWorks)

"While we might tune a system for faster reboots which this technique,
it begs the main questions (why are we rebooting)."

There was a pretty good comment on slashdot on this, the main reason for rebooting seems to be that people who can't/don't want to keep their machines running 24/7 want to shut it down from time to time. This issue is the biggest with laptop users, who ofcourse can use suspended mode, but even that is out of the question in most cases.

ofcourse, i can't see the issue with boot times at all, i got used to my old P166 booting W2K, and it took somewhere around 15 mins from power on to reach usable state.

(Log in to post comments)

Faster is nice but ...

Posted Sep 20, 2003 23:38 UTC (Sat) by euvitudo (guest, #98) [Link]

I currently have two laptops: one for personal use, one for work. I can suspend the personal laptop just fine, and do not have to reboot it. However, since I currently have no battery, I have to reboot if I need to move it across the room, or unplug it for any reason. However, this is really not a problem.

The real problem I have is with my work computer. It has fairly new hardware (February/March 2003), which is the main problem. APM works, but when I wake up the system, it is frozen, and totally unusable. ACPI seems to work somewhat, but I still have problems with it freezing. This is due to an integrated mini-pci wireless card, and/or the Radeon 9000 that X can't seem to bring back up. Quite annoying really. Hence, I have to reboot it when I travel to and from work.

Therefore, faster reboot times are relevant and could be very useful to me--at least until ACPI is supported for my machine.

Cheers!

Faster is nice but ...

Posted Sep 22, 2003 0:58 UTC (Mon) by gdt (subscriber, #6284) [Link]

Reboot time matters in telco and control applications. Watchdog timer fails and reboots OS uncleanly. That eats into the 5 minutes of outage per annum implied by 99.999% availability.

Of more value to the typical distribution is the listing of dependencies in the init script. No more guessing at init seqeunce numbers.

Note that parallel starting of daemons can also make the machine boot slower (eg, if daemons fight over disk access). So dependencies need to be a bit more refined than suggested by the article).