I'm at work now, could someone throw a test with this command?

Nevermind, customer just left.

it works. Thanks

Running with Profile:

Starts: 12:00 every day.
Runs the script at start.

Ends: 18:00 every day.
Runs the script at end.

I'm monitoring my battery and applications from now on, will tell you if there's any noticeable battery drain, sluggishness and application errors.


Thanks to everyone this far!

Cached memory *is* essentially free memory. It's only used to speed access to something which is no longer needed right now, but might be needed again in the near future. If "real" memory is required, then that cache will go away.

Hello my fellow maemo.orgers.

I did keep the profile running for yesterday and the outcome.

Idle consumption did not change.
No sluggishness noticed
No errors.

I will try to tighten the automated drop cache profile from:

Starts: 12:00 every day.
Runs the script at start.

Ends: 18:00 every day.
Runs the script at end.

to this:

Profile 1:
Starts: 06:00 every day.
Runs the script at start.

Ends: 09:00 every day.
Runs the script at end.

Profile 2:
Starts: 12:00 every day.
Runs the script at start.

Ends: 15:00 every day.
Runs the script at end.

Profile 3:
Starts: 18:00 every day.
Runs the script at start.

Ends: 21:00 every day.
Runs the script at end.


Could someone try the same and post some results too?

I'm running stock CPU clock, pr1.3 and FasterN9 with ProfileMatic & Drop Cache script mentioned earlier in this thread.


EDIT:
(12:00 & 18:00 auto-free memory profile)
Battery 88%, Lasted for 4 hours, 57 mins
Running with rules: Silent, BlueTooth off, GSM mode, Availability offline and Power saving when idle for 30minutes.

Power usage:
Active - 12mins - impact 45%
Idle - 4h 47minutes - impact 55%

Avg consumption: 19mA
avg active: 198mA
avg Idle: 11mA

Estimated usage time left: 2d 19h 37mins
Estimated active left: 4h 34mins
Estimated idle left: 2d 2h 28mins

Backlight: 80% - 30seconds.

Consumers:
Device idle: 33%
Meego system: 16%
Battery Usage: 11%
Cellular Network: 10%
Audio PLayback: 8%
Clock: 6%
Stand-by Screen: 5%
Home Screen: 4%
Uptimer: 3%
ProfileMatic 2%
Phone: 1%
Synchronization services: 1%

Those that are bolded seem a bit weird. Any idea why Clock for example is hanging on that list? And why Battery Usage app eats 11%?

The most exciting thing about Drop Cache for me is that it is really able to solve some 'slow visual performance' issues on N9 sometime. (I am the author, and an originally primary target of the app was mostly to do some performance tests/measurements only).

I did not believe to users feedback (I thought it works like placebo pills) before I have tried it with a very sluggish animation between Home screens (I have closed every application before - without any improvement). 'Drop Cache' somehow solves such issue.

From my point of view - there is some low level bug somewhere in the kernel, or perhaps some wrong configuration for cgroups, or something like that - and it is just resetted by such 'drop cache' action. It would be nice if it would be possible to do such reset without full cache reset - perhaps I will try to dig in deeper later. (But it is really hard to reproduce the problem).

Yes, more battery consumption at idle state, even some visible delay if user is doing something (like playing in a game) for short period of time, and so on.

Anyway, I will try to update the application later, with some warning to user, perhaps do it at idle state + charging state only, and so on.

@mdn,

I don't have an N9 so I cannot check. What's the (default) value of /proc/sys/vm/dirty_background_ratio?

Perhaps the N9, given its huge available memory, has a large percentage of memory available for the write-cache, so that when it actually has to write everything to "disk" the whole system halts for a while.

The N900 has ratio = 10%, which is the default in the kernel.

The default value I have is 10.

Have you guys tried fastern9? Speeds up the n9 to no end. I'm 99% lag free.

It is 10, dirty_ratio = 40.
It should not be related directly as mostly helps just a dropped cache - without 'sync' in advance.