Messages in this thread |  | | From | (Alan Brown) | Subject | Re: The best filesystem for a BIG news server | Date | 5 Nov 1996 03:46:54 +1300 |
| |
In article <55jgip$qt9@miriam.fuller.edu>, Christoph Lameter <clameter@miriam.fuller.edu> wrote:
>Note that we are continually swapping drives. No SCSI drive has survived more than 8 Months.
CHECK THE DRIVE CASE TEMERATURE!!
Check the ambient temperature inside the case too.
> We are now checking out how the >situation is with IDE drives for a newsserver spool.
IDE drives won't survive as long as SCSI, mainly because the cheap ones aren't engineered as well as scsi drives. (good, fast, cheap, pick any two)
Keeping drives cool is the secret. If the room's ambient temperature is 20C, the temperature inside a case with average airflow and a few drives installed can quite easily reach 40C, and the drives themselves may well exceed 60C.
At temperatures like that, lifespans are measured in months, sometimes weeks. Remember that every 10C increase in semiconductor junction temperature halves the device's operating life, and much the same applies to drive bearings (platter and servo).
Seagate won't warranty a Barracuda which hasn't been installed in compliance with their required airflow specifications. The number of failed 4Gb units I've seen which were in tiny boxes with no airflow is astronomical, whereas ones installed with decent amounts of air flowing across them last for a few years. It just goes to show that people never RTFM, even when doing so would have saved them a few thousand dollars in time and replacement costs.
There's a reason the old-style mainframe rooms were kept cold enough to require a woollen jumper to work in them for any length of time. It's a pity more people setting up PC based servers don't realise it applies as much now as it did 20 years ago.
AB -- If you go down to the woods today you're sure for a big surpise: Teddies in leather who bang heads together, with black liner round their eyes.
|  |