Re: [PATCH 4/4 v2] Add documentation for hard disk shockprotection interface

[Elias Oltmanns <eo@nebensachen.de>]

raz ben yehuda <razb@bitband.com> wrote:
> On Fri, 2008-09-19 at 11:50 +0200, Elias Oltmanns wrote:
>> "Raz Ben-Yehuda" <razb@bitband.com> wrote:
>> > Elias
>> > looks very interesting.
>> > 1. How will it impact VM , swappers , raids ?  
>> 
>> What exactly do you want to know? As long as the disk is parked, you can
>> do no I/O on it, i.e. no swapping is possible. Since I don't know
>> anything about the implementation of raids and don't have a raid setup
>> here, I can't really tell how a raid device will behave exactly when one
>> of its disks has been parked. Generally, all I/O will be delayed until
>> the disk is unparked again.
>> > 2. you said it has no value for servers ? did you try ?
>> 
>> Again, I'm not quite sure what you want to know. Obviously, desktops or
>> servers for that matter can have disks installed that support the unload
>> feature. On these devices you can use the interface as described in my
>> document. All I'm saying is that the interface has been designed with
>> laptops in mind because I don't see in which situation immediate head
>> unloading would be required or even desirable in a classical server
>> environment.
> I will rephrase. does head parking increase reliability ( MTBF ... MTTF ) of a disk ? 
> My servers hardly generate IOs at night times, does head parking at nights extend 
> the life of a disk ? or is it the opposite ? does it matter ?

Disk head parking is meant as a short term emergency resort. In general,
it does *not* extend the lifetime of a disk. In fact, emergency head
unloads are considered to be costly and smart even counts those events
as an aging factor. It really is just a means to prevent even worse
damage due to heavy shocks, nothing for regular maintenance.

Hope that helps,

Elias