> > I want the
> > swap pager to *first* page to pages resident in another machine's
> > memory. The premise is that network transfer speeds exceed disk
> > access speeds, and are rising at a greater rate (especially on the
> > commodity hardware involved).
> 
> I can understand what you are trying to do.  However, limited cpu
> resources could invoke process migration in addition to limited memory
> resources.
> 

This is the premise for what my clustering project works.  It is the
idea of extending SMP to be SMP over the net.  That is to say load
balance processes across multiple CPU's.  Sort of like NUMA (Non
Uniform Memory Access architecture, where a bunch of machines are 
bolted together with a low latency interconnect and one OS runs across
all of the machines.) except that the OS would no longer be the single
point of failure (other issues arrise though about coherency.)

I have socket migration worked out so that a process can move from 
one machine to another and keep open sockets working, but processes
attached to shared memory have been a limiter.  

I had thought about setting up an alternate page mapping scheme that
that has an extension from just swapped in or swapped out, but swapped
to another machine.  

I am hoping that the season is more ripe for this because when I was
asking questions some months ago I got no responses from the illustrious
members of this list.  I hope I do better this time... :-)


Andy Sporner