This is what I use, I don't need to monitor it or anything, it is totally automatic.
1st. Have "CGwatcher" set to run on startup then run cgminer.
2nd. Set CGwatcher to restart sick or dead GPU's, plus restart if hashrate falls below X.
3rd. Install USB Watchdog Card on Rig. - http://www.berkprod.com/Product_Web_Pages/usb_v1_pc_watchdog.aspx
4th. Plugin USB watchdog to rig( USB connection).
5th. Run USB watchdog "tickler" software, set it to monitor cgminer.exe every 2 seconds.
6th. Test operation of rig before connecting the Watchdog card's reset pins to the motherboard, Have cgminer running, then quit it, watchdog card will beep relay click and the motherboard will be reset.
Computer will restart upon the reset, CGwatcher will start up and start cgminer.
If for any reason cgminer fails, the motherboard gets reset.
If a gpu dies or goes sick and cgwatcher tries a cgminer restart and the computer/cgminer hangs, crashes or freezes, the motherboard will get reset and the rig will reboot..
Plain and simple anything crashing cgminer or stopping cgminer and the motherboard is reset.
Each rig monitors it's self, only input from you is checking hashrate now and again to makesure all gpu's are healthy and not on their way out.
Simple and effective.
One more time. This watchdog card will not work on h/w hung systems. Reset of the motherboard does not work in those situations. Only a power cycle does.
cgwatcher will not detect AMD driver crashes on Windows. My akbash watchdog does.
Only external controller works. Another cheaper solution is to use rpi and control 16A/20A/30A relay via GPIO. You can get these relay boards (optocoupler+relay) on ebay for <$10 and hook them up to your rpi.
One GPIO port for each relay. GPIO drive the optocoupler (3-5mA) directly. You can control 8 or more computers from one rpi.
Or pick this one: http://www.opto22.com/site/pr_details.aspx?cid=3&item=120D25
it will draw ~2.3mA from the GPIO, optocoupler, solid state. Reliable for life. Can be reused for other home automation projects.