Further memory question [was: Clear the computer's memory?]

[R Kimber]

On Thu, 30 Mar 2006 16:46:19 +0200
Alan McKinnon <alan at linuxholdings.co.za> wrote:

> When you say "free", where do you get the figure from? I'm assuming a 

'Free' as reported by Gkrellm.

> 2G machine and after everything starts up it takes about 200-300M. A 
> day later, 1200-1300M is in use. Well, that's about right - the 
> kernel grabs chunks of memory that are not in use and uses them as 
> disk buffers. This is OK, as if you start two vmware sessions needing 
> lots of RAM it flushes the buffers and makes the RAM available. 
> Meanwhile the kernel doesn't necessarily release memory used for 
> buffers after they are synced - it might grab completely different 
> pages for the next buffer and hold on to the previous one.
> 
> The kernel manages memory well, so well that you can't easily tell 
> what's really going on. It's not obfuscation, it's just very dynamic. 
> (You know all about it when you run out of RAM and swap, but the 
> kernel can't prevent that for you.) It does it this way because a 
> smart kernel will actively work to prevent heap fragmentation, make 
> efficient use of caching techniques, and a host of other stuff that 
> I'm clueless about :-)

Are you saying that it takes maybe 24 hours for the kernel to work out
which memory it can use for buffers and whatever?  I assumed that it
would do this straight away and just dynamically release memory as I
start up apps.

After boot up and my starting a few small apps, Gkrellm says circa
1720MB free.  So I take it that the kernel isn't, then, using that
memory for buffers, but that it grabs about 1GB of it for buffers etc
after about 24 hours ???  That sounds a bit odd.   If after 24
hours 700MB is 'free' why doesn't the kernel use that bit for buffers
too? There is always about 700MB 'free'. What make it ignore that part
of memory, which presumably I am not using? Or is the kernel limited as
to how much memory it can dynamically manage?

Still puzzled.

- Richard.
-- 
Richard Kimber
http://www.psr.keele.ac.uk/