On Fri, 12 Feb 2010 16:56:48 +1000 Shane <ibm-m...@tpg.com.au> wrote:
> For those interested, here is the discussion on lkml; > http://lkml.org/lkml/2007/1/23/263 > > It's not like it was rejected without any consideration. > In the time I have been (occasionally) lurking lkml, this has been > raised more than once - and got nowhere. > Maybe Alan Cox will offer a comment. The kernel takes the view that free memory is waste memory. The page cache therefore isn't turned into free space as pages are not needed but instead the vm makes sure most of it is 'clean' - ie correct on disk. That means that when pages are needed the vm simply tosses stuff out of the page cache. If you suddenely need a lot of memory then all the unused bits in the page cache get tossed. When measuring it also remember that a lot of stuff in the page cache will be things like undirtied pages of executables you are running, mapped files etc - it's not a simple buffering system for disk writes. Virtualisers normally interact with all this by having drivers in each guest that allocate memory in respose to global hypervisor managed demand and peeking at local behaviour. The allocated pages are then removed from the guest and given by the hypervisor to another guest and vice versa as demand shifts. Effectively the total pages available is rather lower than the guests perceive. Alan ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
