> 3. System-wide (i.e. global) checkpoint. In this design, the entire machine is
> periodically snapshotted and written to disk. This is the EROS and KeyKOS
> design. It can be done with very low overhead.
I am dying to see an efficient checkpoint of a 300Gig usenet object spool which is receiving articles 24x7 at about 150 a second, shared with 12 servers feeding about 8000 clients in parallel, feeding a filtered subset of the stream to a collection of remote sites and expiring objects at a similar rate to arrival.
That is the challenge.
I accept without reservation the snapshot the entire box model makes sense for a simple appliance with little state. Thats an established old old method. (example: ADVENT - the original fortran monsterpiece)