Maker Pro
Maker Pro

[semi-OT] Harddrive spinup delay for IDE drives

C

Christoph Loew

Jan 1, 1970
0
Hi everyone,

I've used 3ware IDE RAID-controllers to create two
fileservers (1 TB and 1.5 TB low-cost backup) and
while copying data from one 8-disk RAID array
to another 6-disk array i noticed that my PSU
for that server (650 Watt) was unable to supply
power during spinup with all 14 disks (plus 2 Athlon
2200 and some small stuff) connected to it.
I got it working by using a separate PSU for the
8-drive array's drives but in a future extension
the 650 Watt PSU is supposed to drive 12 disks
in the same server and i'm worried about reliable
spinup.

Since the ECC-RAM initialization takes ca. one minute
I was thinking of splitting the 12 drives in two sets
and powering the two sets after a 15 and 30 second delay
using some FET's and two NE555 to delay +12V/+5V powerup.

That would result in this boot sequence :
00 seconds - power to mainboard
15 seconds - power to harddrives set 0
30 seconds - power to harddrives set 1
60 seconds - ECC is initialized, system boots

Has anyone done something similar successfully ?
Unsuccessfully :-/ ?

Are there some power-up standards for PC's I'm unaware of ?


Chris
 
R

Rich Grise

Jan 1, 1970
0
Hi everyone,

I've used 3ware IDE RAID-controllers to create two
fileservers (1 TB and 1.5 TB low-cost backup) and
while copying data from one 8-disk RAID array
to another 6-disk array i noticed that my PSU
for that server (650 Watt) was unable to supply
power during spinup with all 14 disks (plus 2 Athlon
2200 and some small stuff) connected to it.
I got it working by using a separate PSU for the
8-drive array's drives but in a future extension
the 650 Watt PSU is supposed to drive 12 disks
in the same server and i'm worried about reliable
spinup.

Since the ECC-RAM initialization takes ca. one minute
I was thinking of splitting the 12 drives in two sets
and powering the two sets after a 15 and 30 second delay
using some FET's and two NE555 to delay +12V/+5V powerup.

That would result in this boot sequence :
00 seconds - power to mainboard
15 seconds - power to harddrives set 0
30 seconds - power to harddrives set 1
60 seconds - ECC is initialized, system boots

Has anyone done something similar successfully ?
Unsuccessfully :-/ ?

Are there some power-up standards for PC's I'm unaware of ?
Not any that I'm aware of, but there is one thing that springs to
mind - hold the system reset line low until all the disks have
spun up; that way you can bring them up at your leisure (so to
speak) and when you release the reset line, you've got a known
stable system, hardware-wise.

The idea of sequencing _all_ of the power in the right order keeps
nagging at my headbone - it would be a good idea to look into that
as well. (sorry I can't be any more detailed, but I let the smart
people take care of that part. :) )

Good Luck!
Rich
 
P

Pooh Bear

Jan 1, 1970
0
Christoph said:
Hi everyone,

I've used 3ware IDE RAID-controllers to create two
fileservers (1 TB and 1.5 TB low-cost backup) and
while copying data from one 8-disk RAID array
to another 6-disk array i noticed that my PSU
for that server (650 Watt) was unable to supply
power during spinup with all 14 disks (plus 2 Athlon
2200 and some small stuff) connected to it.
I got it working by using a separate PSU for the
8-drive array's drives but in a future extension
the 650 Watt PSU is supposed to drive 12 disks
in the same server and i'm worried about reliable
spinup.

Since the ECC-RAM initialization takes ca. one minute
I was thinking of splitting the 12 drives in two sets
and powering the two sets after a 15 and 30 second delay
using some FET's and two NE555 to delay +12V/+5V powerup.

That would result in this boot sequence :
00 seconds - power to mainboard
15 seconds - power to harddrives set 0
30 seconds - power to harddrives set 1
60 seconds - ECC is initialized, system boots

Has anyone done something similar successfully ?
Unsuccessfully :-/ ?

Are there some power-up standards for PC's I'm unaware of ?

If it was SCSI, I believe that the controller card could do this (
spin-up delay ). :)

Graham
 
C

Christoph Loew

Jan 1, 1970
0
Pooh said:
If it was SCSI, I believe that the controller card could do this (
spin-up delay ). :)

That's right.
However, the difference in price between 1.5 TB as SCSI and as IDE is
.... well ... non-negligible ;-)

Maxtor's 300 GB drives spin up immediatly after powerup and even if
the controller delays and staggers initialization (which the 3Ware
ones do quite nicely) that doesn't help.


So far, the suggestion of applying reset until the forced staggered
spinup is finished seems to be the best idea - but I'll wait until
i have some sacrificial hardware around before I try implementing that.

Any further ideas/comments/experiences/pointers ?

Chris
 
Top