[ubuntu-uk] Load Unload Cycles

[Tom Bamford]

Hi Dougie,

Thank you for explaining it a bit more. I have read documentation from
Hitachi on the subject, but couldn't find anything useful on the Samsung
website. I've also noticed today that the count is going up just as
quickly when running on external power as it does on battery power, so
whatever it causing it either doesn't realise when I plug in the machine
or is doing it for a different reason. It has risen today to 360,563 -
an increase of nearly 5000 since last night! I've applied the patch and
it seems to have stopped instantly; I had to set the drive APM parameter
to 254 before it had any effect.

I realise that the cycle count may not even have an effect on the
drive's lifespan, but I use my machine for about 10 hours a day and
leave it powered on the rest of the time. , however I think I'm going to
have to accept that my drive just won't make the 5-8 years it was
designed for.

Thanks again,
Tom


On Thu, 2007-11-01 at 14:05 +0000, Dougie Richardson wrote:
> Hi Tom,
> 
> Samsung quote the load unload cycle threshold as 600000, so in your case
> with such a high number of counts I'd be inclined to apply the
> workaround.
> 
> There is a correlation on some drives but it depends on the
> manufacturer. Hitachi and IBM use ramp or rollers to lift the heads from
> the disk rather than impact on a landing zone.
> 
> As I understand it from Samsung's documentation, they use landing zones
> or component start/stop zones. The idea is that when the drive powers
> down the heads are landed on an area usually in the centre of the
> platter that isn't writable, hence avoiding corruption.
> 
> It also isn't as simple as failure at the quoted threshold. The actual
> figure is attained by testing and shows the minimum number of hits
> landed before the chance of damage reaches 50% - in other words beyond
> this threshold damage may occur but below it shouldn't.
> 
> The problem is that in a drive to improve power efficiency has caused
> drives to be powered down more often, increasing the amount of counts.
>