In article <4pp6sg$8v2@fg70.rz.uni-karlsruhe.de>,
ig25@fg70.rz.uni-karlsruhe.de (Thomas Koenig) writes:
> Hmm.. let's not forget real swapping, where a process which consumes
> loads of resources is swapped to disk totally. (Question: Could this
> be accomplished by simply stopping it for some time, and let 'natural'
> paging do the rest?)
It could be accomplished also by decent scheduling of the batch jobs.
Either way, I'd probably rather see this done in user space, perhaps
by having a batch daemon which decides which jobs deseve to be swapped
out when. Some kernel support might be necessary, but using standard
job control --- sending SIGSTOP and SIGCONT to schedule the running
batch jobs --- might be sufficient.
> Also, preemptive paging would be great (this means writing copies of
> active memory pages to disk, so that they can be thrown out of core
> without writing them to disk first if more memory is needed).
Planned for 2.1. There are a number of places where I want to
improve swapping throughput further, especially to do with even better
clustering of writeouts and managing sequential access of large
datasets more intelligently. NFS swapping should also appear at some
point.
Cheers,
Stephen.
--- Stephen Tweedie <sct@dcs.ed.ac.uk> Department of Computer Science, Edinburgh University, Scotland.