Disk noises and Tyan S2468UGN

Mark Hahn hahn at physics.mcmaster.ca
Sat Sep 14 08:34:31 PDT 2002

> > http://www.storage.ibm.com/hdd/ipl/oem/tech/pfa.htm
> >
> >A (sic) periodic intervals, PFA's Generalized Error Measurement (GEM)
> >automatically performs a suite of self-diagnostic tests which measure
> >changes in the disk drive's component characteristics."

which is a GREAT feature.

> So I read the URL and don't see what this feature buys you on Linux

why do you think there's some OS specificity?  afaikt, the results 
would be reflected in SMART, which is most definitely available under Linux.

> Even if PFA is a plus, why must gathering the information
> require the generation of such an awful sound?

seeking is inherently noisy.  
fast seeking is inherently noisier than slow seeking.

many (most?) drives nowadays have an alternate mode that trades 
slower seeks for quieter operation, though its common to turn this 
off for obvious reasons.  but if the disk is trying to measure its
own seek performance, it's not too surprising that it tries fast seeks.

I'm guessing that the only thing that needs to be tuned here is 
how often the drive looks for measurements.  obviously, if you keep
it fairly busy doing real work, it shouldn't need to do PFA-specific
seeks nearly as much.  similarly, if it's idle, it should probably
do unsolicited seeks fairly infrequently, since there's little chance
of spontaneous degredation.

More information about the Beowulf mailing list