Messages in this thread |  | | From | (Jason McMullan) | Subject | Re: Core Dumps & Restarting | Date | 1 Nov 1996 03:55:27 GMT |
| |
Timothy Peters (tim.peters@nene.ac.uk) wrote: : I am currently investergating the posability of a research project in : this area (Resistant Storage in Operating System Recovery).
: I will be using linux as a base for my work which will involve try to : get a system to halt and restart as if nothing had happend, and also the : possability of creating two indpendant machines that act as one so that : if one crashes the other can pick up and continue with out the : applications noticing.
: If any body has any idears of suggestions or knows of any other work : that is being done in this field please let me know.
First off - don't use Intel Linux - use Sparc Linux (or even Atari Linux) instead. Intel machines are too hardware happy (read 'unknown hardware states') for that to work.
But some issues you will have to address: * SIGALARMs that are late (and a whole slew of other timing problems) * Cache flushing issues (what was MMAPed, what was in swap, etc) * Dropped network connections (an almost impossible problem w/out _very_ special hardware to keep alive TCP/IP connections) * Unknown hardware states (I lost power in the middle of getting a packet from the SCSI HD - do I return an error on bootup, or retry? - repeat for _all_ devices)
I wish you luck... This is a PhD thesis, right? (you'll need the years... or a lot of grad students)
Jason McMullan - Research Programmer, Robotics Institute, CMU
Me: http://www.ul.cs.cmu.edu/~jmcc Linux GGI: http://synergy.caltech.edu/~ggi
|  |