Re: Resource Allocation Strategy

Hi Neal,

"Neal H. Walfield" <[email protected]> writes:

> Instead, we emphasize decentralized policy specification and
> centralized computation of schedules based on these policies.

Isn't this similar to what you were heading to with your
L4/physmem/libmm experiments?  I.e., `physmem' is the one server that
"schedules" physical memory, but applications implement their own policy
and have a way to tell `physmem' about their preferences[*].

Thanks,
Ludovic.

[*] This was also the topic of your talk at LSM 2005.