Alexander Dalloz wrote:
These drives were very short lived and usually died out within 200 days for the first onset and others had a little longer lifespan near 300 days with none making it any further without failures.Am Mo, den 13.06.2005 schrieb Jim Cornette um 3:43:
This program should be sufficient for the purpose. Is 46 Degrees C hot for a light duty installation? The drives in the environment that they are used are constanlty performing read/write operations and are powered on 24/7
Jim
Installed: hddtemp.i386 0:0.3-0.4.beta13 Complete!
hddtemp /dev/hda
/dev/hda: IC25N040ATCS04-0: 46°C
46°C is too much in my opinion. It is not that bad that you very quickly will have problems but the temperature will massively lower the drive's lifetime. I suggest you do a better cooling. Place a 80mm fan in front of the harddrive(s). That helps a lot. 30°C - 35°C are values which mean a good condition.
Alexander
My plan is to set up two identical machines as the product usually is sold as and run the tests with one using the now no longer used Hitachi, which smokes at low levels of duty compared to the Fujitsu, which seems to be running reliably well over 500 days of continuous service.
One scheme that was tried was to put a fan to blow air directly at the hard disk. The scheme was tried on around 20 computers and the machines did not fails as of yet. Before this scheme, the only fans were the CPU, power supply fan and another case fan to cool the whole unit.
Other than the 20 test units, the remaining units have the above configuration.
A newer scheme uses larger drives which consume twice the power. (changing from 2.5" to 3.5" drives and raising the profile of the case. No additional fans added)
I didn't think the changing of drives to 3.5", though larger platters would have more surface area to throw out the heat, would be a better alternative over a better and a cooler running 2.5" drive.
Adding an additional fan and better cabling seemed to be a better route to take, along with good drives used in the first place. Some "frankencables" were used in these units also. (40-pin IDE cables with 2.5" to 3.5" adapters used). With the 20 test units, the "frankencables" were replaced with 40-pin to 44-pin cables. I believe the better cabling was more responsible for the 20 units not making any repair visits, but the extra cooling and switching of vendors for the hard drives all happened within th same timeframe.
I have one of the converted units that I kept. Running this unit against the standard build unit and also testing the 3.5" converted units might be interesting.
So the test can be ran with similar stress fo each unit, what might be good program to run in these units for a moderate to heavy constant load to be delivered to each test computer?
Jim
