> well, 'free' divides it into 'total, used, free, shared, buffers, cached':
> it's the last  3  columns which i don't fully understand.
> 'shared' is always  0 , so it mb used only on multi-user systems;

I've been on Linux systems where shared is not always zero. I would
think that maybe it's not being reported correctly, actually. One
would think shared is any memory that's being used for multiple
things, so for instance libc, since nearly everything uses uses
parts of libc on a shared basis. And Linux is a multi-user system,
even if you're the only user, in which case we have several instances
of you using the system, or the degenerate case of only one user on
the system, which is only during single-user mode. And if you have
multiple users (even multiple isntances of you) you have multiple
instances of shells running, possibly multiple editors, multiple 
getty processes and so forth running. A fair amount (actually,
anything that doesn't change) of that should be shared. As a
rule, then, data and stack  pages aren't typically shareable, since
they change for each user and/or instance. But code pages are usually
shareable, since they don't change.

> 'buffers, cached' seem to make use of otherwise empty space,
> as the 'used' figure is usually close to 'total'.

buffers seem to be self-explanatory, i.e., buffers allocated to
disks etc. cached would be everything else. For instance, if you
plan on doing a lot of compiling, there would be instances where
lots of the same code gets reused - for instance, between subsequent
invocations of gcc. I would think that the system would get those 
pages from cache rather than have to reload pages from the gcc
binary for each compile.

> 


Want to buy your Pack or Services from MandrakeSoft? 
Go to http://www.mandrakestore.com

Reply via email to