Re: currency based virtual memory

[Matthew Sackman]

On Thu, Nov 07, 2002 at 03:00:07PM +0100, Niels M?ller wrote:
> Matthew Sackman <address@hidden> writes:
> 
> > Suppose x people are logged into one machine and are using it with
> > similar workloads. Then, say one person wants to compile an
> > application. With the economic model it would *appear*[0] that what
> > would happen is that in order to optimise the global system, the system
> > would delay that compilation as being too memory exhaustive and too cpu
> > intensive.
> 
> No, that's not what's expected to happen. What should happen is that
> the processes owned by the user that runs the compilation will suffer
> more paging than processes owned by other users. (Of course, there are
> lots of different ways to configure the resources, but that's the
> idea, at least).

Ahh - that's put my mind at rest. I was hoping I'd misunderstood. So the
compilation causes a depression just for that user.

> And when memory gets scarce, the price rises, and each user needs to
> give back unused memory to the system in order to afford memory for
> important processes. 

But other users should not have their processes swapped out or
suspended? Just that if they have memory left over that they're not
using that they got when memory was cheap that it should now be
returned.

> And it's all about the physical memory, space on the swap should
> be free, or at least a lot cheaper than physical memory.

Um, if a process decides to give up all its RAM and buys swap instead
because it's cheaper, how does it manage to buy back RAM when the prices
come down? It's swapped out!

Matthew

-- 
Matthew Sackman

BOFH excuse #114:
electro-magnetic pulses from French above ground nuke testing.