This one should be fairly easy to implement.

 

On top of, or as a user configurable option, allow sequential spin up of drives.

 

The idea being that in normal operation spinning up a drive is the most energy demanding task. By sequentially spinning up drives an artificial spike in the PSU load can be avoided at the small expense of a few more seconds spin up time.

I'm not sure this is possible at the software/application layer. Maybe at a bios layer, power supply layer or software setting in the drive itself.

 

With SCSI drives it was simple, there is a jumper for the drive to only spin up by software command.

 

Found some information here:

http://scottstuff.net/blog/articles/2007/10/22/sata-staggered-spinup

http://www.highpoint-tech.com/PDF/Staggered_Drive_Spin_Up-A_Primer.pdf

http://en.wikipedia.org/wiki/Staggered_spinup

 

I thnk this has the most promising detailed information.

Something called power up in standby.

http://ftp.t10.org/ftp/t13/technical/d97150r0.pdf

 

 

Further research reveals that WD GREEN Drives have a jumper.. This is a very positive find.

http://www.hothardware.com/articles/Western%5FDigital%5FCaviar%5Fand%5FRE2%5FGreenPower%5F1TB%5FHard%5FDrives/

drives sport a set of eight pins that can be jumpered to change the drive's configuration. The close-up of the label above shows you the three jumper settings: jumpering pins 1 and 2 enables SSC (Spread Spectrum Clocking), jumpering pins 3 and 4 enables PUIS (Power Up In Standby), and jumpering pins 5 and 6 enables 1.5GB PHY.

 

And some details from my drives.. I bet we'll get to the bottom if this.

 

parity device:  pci-0000:06:00.0-scsi-0:0:0:0 (sde) ata-ST31000340AS_3QJ08RZW
root@Atlas:~# hdparm -I /dev/sde  | egrep -i power 
           *    Power Management feature set

disk1 device: 	pci-0000:00:1f.2-scsi-1:0:0:0 (sdd) ata-WDC_WD10EACS-00ZJB0_WD-WCASJ1352706
root@Atlas:~# hdparm -I /dev/sdd  | egrep -i power 
           *    Power Management feature set
                [b]Power-Up In Standby feature set[/b]
           *    SET_FEATURES required to spinup after power up
           *    Host-initiated interface power management

disk2 device: 	pci-0000:00:1f.2-scsi-0:0:0:0 (sdc) ata-Maxtor_6H500F0_H80C0QFH
root@Atlas:~# hdparm -I /dev/sdc  | egrep -i power 
        Advanced power management level: unknown setting (0x0000)
           *    Power Management feature set
                Advanced Power Management feature set

 

Further research reveals...

 

root@Atlas:~# hdparm -h | grep standby 
-s   set power-up in standby flag (0/1) (DANGEROUS)
-S   set standby (spindown) timeout
-y   put drive in standby mode

root@Atlas:~# hdparm -s 1 /dev/sdc

/dev/sdc:
Use of -s1 is VERY DANGEROUS.
This requires BIOS and kernel support to recognize/boot the drive.
Please supply the --yes-i-know-what-i-am-doing flag if you really want this
Program aborted

I experiemented with that power up in standby mode - not sure if it was with a Seagate or a WD.  It never spun up - EVER.

Very interesting indeed. I never even considered this being viable at boot up time. This idea was solely to replace/augment the SPIN UP button on emHTTP.

 

If it can be taken further than that then fantastic but the original idea is way less ambitious than that.

Then I did not understand your real request.

if drives are all spun down and you click the spin up button, they are already spun up sequentially.

At least this is how I've seen it happen.

ah! thanks weebo, I thought of writing this (that already when you click spin up the start sequentially) but was afraid that I would say smth stupid

 

 

I experimented with that power up in standby mode - not sure if it was with a Seagate or a WD.  It never spun up - EVER.

There's no guarantee that it works with all controllers.

I tested it on 3 drives, 2 accepted it, 1 did not. However, the controllers did spin up the drives.

The WD green drives work perfectly. (for me at least)

 

Then I did not understand your real request.

if drives are all spun down and you click the spin up button, they are already spun up sequentially.

At least this is how I've seen it happen.

 

Your right, bad wording on my behalf. Currently they are spun up as quickly as they can be. The spin up commands returns as complete way faster than the drive is actually available so what we have is basically a PSU hammer of a few seconds.

 

So what I am really saying is a configurable sequential  spin up delay

I like the idea of delayed spin up.

 

If we could fix this power on issue (which is a 2x spike at power on) and do this also, we could reduce the requirements on our PSUs by 1/2!  Fixing the boot issue seems unlikely unless some clever electirical engiineer started selling a time delayed molex plug the could install between the PSU and the drive.  They could come in 2, 4, and 6 second delay varieties, people could tune their drives to start spinning a few at a time.  For a couple bucks a piece, I'd probably be in for 10-12!

Delayed spin up on boot.

 

The WD green drives have a jumper. Plus if you set them in software, it is remembered.

Also from what I read. If pin 11 is floating the drive is not spun up until the controller sends the command.

In addition, certain backplanes have this pin floating so it expects the bios or operating system to spin up the drive.

 

 

I believe I read that PSUs have their peak efficiency usually in the viccinity of 50-70% max load. So we'd want to be targetting this number.

 

I read another article that puts the 1TB drives at about 8+Watts load, and a little less under IDLE (exception being the WD Greens that have about 4.5Watts at idle).

 

The WD greens not only offer better IDLE power consumption than the competition (with load power consumption being similar to competitors IDLE), but they only require 14Watts at spinup (vs. 30watts for the comparable Seagate).

 

So lets say we build for minimal power draw on the hard drives. 16 WD 1TB Green HDDs will reqiure 224Watts at spinup. Say you use a low power system that draws at max 50Watts you would do well with a 300Watt PSU which operates in peak efficiency around 150-200Watts.

 

16 drives idling is about 72watts and lets say the rest of the computer IDLING is 30Watts and ideally you would want a PSU with a sweet spot at 100Watts. Then factor in that the machine could conceivably have all drives spun down.

 

I wonder if you could build a power supply similar to how speaker amplifiers can have massive headroom for occasional peaks.. that would be nifty. A couple of power reserves for the spikes when needed but otherwise cruising along nicely at the lower end.

 

I'd also like a PCIe X16 card with 16 ports without RAID capability for really cheap prices... someone build one quick!