Automatic means for spinning down disks available?
Gary Kline
kline at tao.thought.org
Thu Apr 12 01:22:09 UTC 2007
On Wed, Apr 11, 2007 at 04:14:33PM -0700, Chuck Swiger wrote:
> On Apr 11, 2007, at 3:32 PM, Gary Kline wrote:
> > Some things to consider (besides powering -down or -off drives)
> > are battery backup system. Don't most UPS systems isolate your
> > servers from the wall-socket?
>
> The better grade of UPSes do exactly that-- they provide "galvanic
> isolation" by using an isolation transformer which has the primary
> and secondary windings completely separated, and ensuring in the
> design that you don't connect the service neutral line to the output
> or load's neutral line. The load can thus either be floating or tied
> to the local building ground. This type of design is known as
> "double-conversion" because they always feed the input AC line
> through the rectifier & DC inverter, using more power but providing
> better PFC and can provide the load with an AC frequency which is
> different than the input AC frequency (ie, they can provide 60Hz
> output from 50Hz input, or vice versa).
Years ago I spent a lot of money for a top notch surge
protector. It still protects everything to this day; and very
well. Now and then I'll find my LAN down, DNS too, obviously,
because of a surge of one sort or another. The power here
(Seattle) is pretty good -- well, except for wind storms {koff}.
But I'm way past due for having the sort of higher quality
UPS that you're taking about. It would be wired to a pipe struck
in the earth. Floating_ground just doesn't cut it. Any models
you'd recommend? How much system installation is required?
I'm CAT-5A cabled. Software, no problem; anything else is.
>
> Cheaper UPSes, which include almost all consumer-grade models from
> APC, Tripplite, etc run in "line interactive mode", which involves a
> self-tapping or ferro-resonant transformer, can adjust the voltage up
> or down within limits, but they do not perform PFC and cannot provide
> frequency conversion, and they pass the neutral line from AC line to
> load without isolation, thus passing common-mode noise through. This
> design is lighter and requires fewer components (an isolation
> transformer is heavier), and does not keep the DC section and
> inverter always under full load, so are somewhat more efficient, but
> cannot deal with frequency drift or significant voltage changes.
Understand, thanks, Chuck. Here (where rubber-meets-pavement
is where *not* to cheap out).
>
> > At what level do hard drives have identical circuitry so that
> >they can be software lower-voltaged?
>
> The boards within a drive family might be identical (WD200BB/WD400BB/
> WD800BB/etc), but they don't deal with under-voltages at all well--
{{ this is what i was afraid of.... }}
> you'll either pull excessive current through the servo and spindle
> motor windings, or perhaps the drive will fail to spin up entirely.
> The spindle motors are designed to spin at the calibrated speed and
So, pragmatically, a drive is either going full-throttle or
it's OFF. ...Hm.
> won't spin at slower speeds.
Somewhere, prhaps at the Gnome shutdown GUI (dialogue?)
it reads: Off, Changed-user, Idle, Power-Off, Reboot,
or whatever. Flame from Gnome/KDE folks to /dev/null, please.
I'm guessing the "Idle" is for the laptops. YEs/no?
Something else to consider here is how much power do the newer
40-60, 200-300GB drives suck up? I don't think the drain is
much compared to, say, 3 CRT television sets drowning on
several hours/day. Still, let's SWAG that there are 25-30
million of us nerd/geek types running at least one computer.
That adds up.
>
> > *Except for consumer __cost__*, why don't all boxes have builtin
> > batteries like laptop? ...There are lots of things to consider.
>
> Cost is the primary reason why boxes don't have built-in batteries.
> People flinch away from paying for real RAID systems which include
> battery-backup for the drives...
Well, then I'm definitively part of the problem; suspect that
most of my kinsfolk are too. aNy idea how mmuch of this could be
solved by software? Maybe when a machine turns itself off at
03:30, it write a state-file. When it reboots {either by magic
timer or by actually crawling around down there and toggling
switches }, presto, you have everything just the way you left it.
puts("Feedback, world?");
gary
>
> --
> -Chuck
>
--
Gary Kline kline at thought.org www.thought.org Public Service Unix
More information about the freebsd-questions
mailing list