rkaiser@sysgo.de wrote:
Hi Robert,
. or am I really on uncharted ground here?
No not completely, if you can live with hierarchical proportional share scheduling, see
Checkout Ford et al. "CPU inheritance scheduling", OSDI 96 http://www.cs.utah.edu/flux/papers/index.html
and for a look at L4 in particular
http://www.cse.unsw.edu.au/~disy/papers/index.html#ug-theses
Simon Winwood Flexible scheduling mechanisms in L4, BE Thesis, SCS&E, UNSW, November 2000.
Have a look at Dresden as well for a more real-time slant on scheduling in L4. I believe they use multi-level round robin and a variation on priority inheritance (helping). I can't convey the exact details of priority assignment etc, so your better off having a look yourself.
I believe Jochen envisaged something "better" than proportional share, but still keeping the flexibility, decomposability, and preserving performance. However, this never got beyond being a vision except for the existence of preempters (time slice + total quanta ideas) in V4, which is unfinished work.
- Kevin