Boot in 5 seconds
 

My N810 takes 33 seconds to boot, while it only loads a few kernel modules and runs a few init scripts.
I think we should bug this:http://lwn.net/Articles/299483/
Booting in 5 seconds.
Shouldn't we?

Re: Boot in 5 seconds
 

well, there's the battery start too. I agree that when its empty and you plug it, it takes a few moments before it decides to boot. But i don't think its all that bad, either, since its not meant to be booted up all that much.

Re: Boot in 5 seconds
 

Hmmm... Sounds like a good goal for Fremantle, doesn't it?

(Realistically, it can't happen before then, because Nokia really doesn't want to jump kernel versions for no good reason. They won't think this is a good reason.)

Re: Boot in 5 seconds
 

http://www.slideshare.net/qgil/osim-...h-presentation

Dunno if someone saw on page 4 the upstart project? :)

http://upstart.ubuntu.com/

Re: Boot in 5 seconds
 

It is 33 seconds without battery start. 20 seconds of blue bar, 13 seconds of Nokia splash screen.
If it can boot in 5 seconds, why not do it?

Re: Boot in 5 seconds
 

Wow, I thought the lwn article was quite interesting. It would be great to be able to boot in five seconds, even without networking.

Couldn't this affect the power-usage of booting and make turning off your tablet a more viable thing to do? It is said never to turn off your tablet, and I only turn mine off when I change batteries, mainly, but it would be nice to get rid of that rule.

Re: Boot in 5 seconds
 

Yup, and at another point (I think the "Core System" talk) it was announced that initrd is going away. It should be relatively easy to streamline Xorg startup (display, keyboard etc config is fairly static) and we don't have a display manager to worry about, so the main missing ingredient is sReadahead.

Well, once you boot you still have to wait several minutes at full CPU while metalayer-crawler eats your battery by re-indexing the contents of your cards. Let's hope the metatracker replacement will be lighter/more intelligent.

Re: Boot in 5 seconds
 

Well, it's hard to guarantee an up-to-date database without rescanning on boot; the user could have taken it out and changed it in another system. I could be wrong, but I don't think FAT stored last mount time for comparison; if other systems can be trusted to change a time-stamp on modify, then a fast scan could be made of all timestamps, comparing each directory to a stored copy made at shutdown. (More generally, this could be applied at every mount/unmount.)

Re: Boot in 5 seconds
 

It doesn't.

I understand the reasons why cards have to be reindexed at every mount, just unhappy with the cost of the current implementation.

Re: Boot in 5 seconds
 

Does Arjen include the GUI? Or was it like XFce?

Matter of( Coreboot,) upstart, faster processor, faster I/O, PowerTOP + LatencyTOP investigation.

Re: Boot in 5 seconds
 

Bugged as #3777

Re: Boot in 5 seconds
 

Please fix that useless summary. . . .

Re: Boot in 5 seconds
 

Suggest one yourself

Re: Boot in 5 seconds
 

Well, that, or we could just close it as INVALID for being uselessly generic. Might as well file a bug to 'Make it fast'. . . .

Re: Boot in 5 seconds
 

You really are a sanctimonious prick, aren't you? I thought you were just playing one, but no...

Re: Boot in 5 seconds
 

I believe there are many possible improvements that can be asked for before booting time, they were listed here on itt, they were discussed throughoutly and maybe we will see some result/answer in the near future (Java support, 3D acceleration for example or support to the built-in image accelerator), but booting time is not that bad afterall and if waiting 33 seconds gives me a stable and working device I can live with that.

Re: Boot in 5 seconds
 

There are people also inside our organization pointing to this 5s holy grail. I will say here what i said internally (not that my word is any authoritative, but I'll still say it :-) ).

********.

There is certainly a point in making the device to boot faster, but look who's talking about it. Intel. Why? Because their current HW sucks in terms of power management, so you _are_ expected to switch it on and off continuously, then of course it is important to not wait forever for X to show up.

Removing modules and going for a single kernel binary? Yeah, right!
This for example would mean that we should start shipping different kernels for n800 and n810. But what is more important, we would have he problem of debugging the thing (you don't want to use in R&D something that is not shipped to production) and go back to the pain experienced till now.

In this sense Fremantle is a huge improvement since we are enforcing drivers to be modular and this allows the creation and usage of tools like the pm-tester i posted about.

If energies have to be put into making the device more efficient, they should be focused on stability, so that the user doesn't have to actually powercycle.

Since 770 we were supposed to be always on, then it became always on, always connected. I see no point in following something that we have obsoleted more than 3 years ago.

Re: Boot in 5 seconds
 

I fully agree with the emphasis on power efficiency and no need for unnecessary rebooting, but since you're talking to existing userbase... will Fremantle even be made available for the current N8*0 hardware?

It'd be nice if Nokia had learned something from the (admittely test case) N770 debacle. With key components still under lock and key I'm afraid there won't be sufficient N8*0 community to keep the N8-generation running Maemo if Nokia diverts all resources to Fremantle (Maemo 5) and the planned N9-series.

Re: Boot in 5 seconds
 

As has been said many times before, it's still unknown. The reality is that current hardware simply may not be able to handle it, but I'm sure Nokia will have a real answer when we're actually close to some sort of alpha-level release of the SDK.

Well, the WiFi driver is open now, so there's not really anything un-solvable holding us to a specific kernel version anymore. . . .

Re: Boot in 5 seconds
 

What kind of answer is that?
There is a goal, boot in 5 seconds, and the summary simply reflects that. I don't know any better how to describe it.
If you don't like the current summary, please suggest one and that will be.
If you think that goal is worth or not, that is a completely different matter, that you can argue at will.

Re: Boot in 5 seconds
 

Instead of looking that it benefits everybody, you are worried because it benefits intel. If that logic were to be followed, then never would be improvements, because they benefit your neighbour.
You can ship the same kernel. The basic idea is to ship the modules you always need in the kernel and keep the rest as modules. And still, what makes most difference, is sReadAhead.
Nice, but... as already has been posted... what about the N8x0 tablets?
Nice goal too, but, unfortunately, power cycles are unavoidable (the watchdog may powercycle simply because there is too much system load).
So, I don't see how having also this goal, fast boot, may conflict with any other goal. I think this goal is worth to keep, and at least see how it evolves in the major distros, and copy whatever is worth to.

Re: Boot in 5 seconds
 

I am stating that, having a finite amount of resources and time, they should be used where they can be more effective.

Which is not boot time.

Even on my laptop i have uptimes of weeks/months so i don't care about boottime because most of the optimization is not involved in suspend/resume to ram.

Re: Boot in 5 seconds
 

No, you missed igor's point: Intel HAS to focus on something like 5s boot time, to take eyes off areas like power management where they don't perform so well.

Every product has its selling point(s).

Re: Boot in 5 seconds
 

Well it was Igor (Nokia) after all who brought up Fremantle in the current context of maacrux's N810 booting in 33 seconds.

FWIW, I fully respect Igor's work and presence here on ITT, but general Linux advancements like in the case of the 5-second boot needn't even be tied to Maemo 5 (Fremantle) and its planned fancy GUI, unless the N8*0-series is facing uncertain future without any continued developer support by Nokia.

Near-instant booting would naturally be wonderful, but I'd be happier if the N8*0 platform remained properly supported for another year or two. :p

Re: Boot in 5 seconds
 

I'd also like to chime in on the whole "5 seconds as the goal" bit.

Plucking a number that sounds good from thin air and then making that a goal can lead to serious unintended consequences. A little better to shoot for % improvement, although even measures like that have to be re-evaluated and retuned with every iteration due to diminishing returns.

I'd rather the team just focus on removing any and all inefficiences where possible and practical, and let the boot seconds fall where they may. ;)

EDIT: hey igor, I'm on YOUR time now-- go to bed! :p

Re: Boot in 5 seconds
 

Sleep is overrated :-D

And anyway I still have some work to do, but if you are in Ruoholahti this week, feel free to drop by, A504.

Re: Boot in 5 seconds
 

Let's suppose that is correct. They make 5 s boot time happen to keep eyes off power management, then every distro moves to 5-10 s boot times and then... we are looking at power management again. That could buy them a few months at most. Ah, and those who refuse to boot fast will look bad (a non-selling point?).
Fast boot is still in its infancy, but soon will mature and will be everywhere (yes, many people turn their computers and devices off when not in use), so my point is to keep an eye on it and make use of it when it is ready.

Re: Boot in 5 seconds
 

Whether it's needed or not (should not!) - fast boot (and in particular a 5s boot) would be really impressive, and it would be a sign of intellectual power :cool:

Re: Boot in 5 seconds
 

I agree that this isn't the right priority -- for Nokia.

Hopefully though, people who like this will be able to submit improvements that help with boot time before the Fremantle release. It harms nobody, helps some people, and makes some people just plain happy about the coolness; if some of the people for whom it is a high priority can make it happen, everyone wins.

Re: Boot in 5 seconds
 

That would be proof that there is a lively community.
This sort of activity doesn't require any information unavailable due to NDAs.

Re: Boot in 5 seconds
 

There's no question it would be impressive...but the point is that it would be completely useless. Considering the waste of development resources to implement...it wouldn't even be useless; rather it would be completely counter-productive and set the platform behind.

You want the impression of fast boot? Uncomment the sections in /etc/systemui/systemui.xml and choose the "Soft poweroff" option. Thanks to G.Antilles for the tip, even if he is a sanctimonious prick.

Re: Boot in 5 seconds
 

Thanks! I'm hopefully meeting TK for supper tonight.

Re: Boot in 5 seconds
 

The LWN article is a little sparse on details, but on this subject it says "The kernel has to be built without initrd, which takes half a second with nothing in it. So all modules required for boot must be built into the kernel" (emphasis mine).

It looks like we're already there on the "required for boot" front, and we now know that initfs is going away in Fremantle (probably for other more important reasons, but if it happens to improve boot speed at the same time no one's going to complain - same goes for upstart).

More importantly, the current kernel on current hardware takes less than 2" to boot (before initfs kicks in) so there seems to be little point in spending any effort specifically to optimise it further.

Agree 100%.

If anyone cares enough to try it may still be useful to see bootchart output to identify any obvious userland bottlenecks though.

Re: Boot in 5 seconds
 

I think I saw some bootcharts recently in these fora, actually... I don't recall what the point was, nor what thread.

Ah, maybe this was it: http://www.internettablettalk.com/fo...087#post223087

Regarding deblet, not Maemo, so not necessarily enlightening.

Re: Boot in 5 seconds
 

Hm, on second thought, will HSPA-capable devices require a power cycle to change the SIM? (I don't necessarily expect an answer, but it's something to consider.)

Hopefully not, but AFAIK all current Nokia devices with a SIM slot do. If so, boot time and power consumption during boot/user session startup suddenly become much more important.

This is one of a very few areas where Nokia should take hints from Apple IMHO.

Re: Boot in 5 seconds
 

Um, you're planing on swapping your SIM card more than once a day? :\

Re: Boot in 5 seconds
 

http://www.cs.helsinki.fi/u/jvtmuukk/chinook/bootchart/ has ones for chinook.

Re: Boot in 5 seconds
 

It IS generic. See https://bugs.maemo.org/page.cgi?id=bug-writing.html , especially "Be specific. The quicker the engineers can isolate the bug to a specific area, the more likely they'll fix it."
Feel free to propose explicit technical changes.

Re: Boot in 5 seconds
 

..and Apple has to add removable battery. Besides that, Apple has SIM lock, and sell it like that as well.

We need to keep in mind the reasons one would want to (or must) reboot.

1) User adds/removes physical hardware not supporting hotswap
2) User wishes to reboot to new kernel
3) User puts the device off because it won't be used for quite a while
4) User starts it up because it has crashed
5) User starts it up because it was previously out of power
6) User used suspend or hibernate and this went broken

Case 1 happens far more often on a desktop or cheap server than in quality servers, laptops, embedded devices (including NIT), and such.

Case 2 happens far less on an embedded device as the NIT than on a laptop or nettop. It only happens when a security or reliabability patch has been added. For NIT, it basically only happens during SSU, and then its expected (right? at least from my point of view it is).

Case 3 happens during travelling; planned usage.

Case 4 shouldn't occur, it destroys user's trust/reliability for/on device...

Case 5 occurs quite often. More often than on a desktop or server, but perhaps more on par with a laptop or nettop?

Case 6 isn't supported well on the NIT while it is good supported on laptops and nettops.

If you add it all up, + and -, there is a slight less importance for fast booting on a NIT than on other devices, except in rare cases.

But let me say this. 33 seconds is acceptable to me. 33 seconds is acceptable waiting for my DVR to boot. 15 seconds is acceptable for my TV. 10 seconds is OK for my radio. For my NIT, 1 minute isn't acceptable. That is the point. I don't like to wait for my laptop to boot a minute either. That too is Intel market, but not Nokia market. And, 10 minutes for GPS (or more, or days...)? No that isn't acceptable at all.

So while 33 sec is acceptable, 25 seconds is even better, and 15 too. It isn't a deal-breaker for most people, but its very nice if it boots quicker. That is why on 3rd page Texrat message makes most sense: try to make the device faster, but don't make it some kind of goal to make it 5 seconds. That is undoable. And, learn from others who are aiming at this.

Re: Boot in 5 seconds
 

Not on most days, but I do tend to travel a lot and EU roaming charges are still not low enough.

Maybe I'm a bit spoiled, but right now I can take the tablet to the airport, browse/check email/etc while waiting to board, play 2-3 hours' worth of music and/or video to alleviate flight boredom and still have enough juice left after I land to check email and fire up the GPS to find my hotel/meeting/whatever (and a power socket).

Rebooting eats around 10% of my battery charge once metacrawler etc have settled, so I'd rather not have to do it :-/