Re: PATCH: flags, paging guest kernel, various

[Kevin P. Lawton]

Ulrich Weigand wrote:

> Well, yes, but the really difficult thing is that the swapping strategy
> should normally depend on the memory pressure that the host OS experiences.
> I.e. the host swapper tries very hard to free pages only if there is currently
> large demand for memory.
> 
> Our algorithm isn't aware of the host memory demand, however.  This means
> that we either agressively prepare lots of pages for swapping, at the cost
> of frequent monitor page faults, even if the host memory pressure is low,
> or else we try to keep guest pages locked longer, which can cause out-of-
> memory situations for the host although we *could* free pages if we knew
> about it :-/

Can we not find this info from most host kernels?

-Kevin