Reducing power consumption of FreeNAS 8.3

[darkryoushii]

Okay, so this:



is your system? Those are 7200RPM drives and Toshiba rates them around 7W active, 4W idle, 1W sleeping. Five of them active is going to be 35 watts, idle will be 20 watts. Sleeping, 5 watts. But many here will tell you, spinning drives up and down is extremely stressful on drives, and will likely lead to unhappiness. It might not be a tradeoff you want to make.

I'd love to spin them down because I don't believe anyone who says that it puts unnecessary wear on a drive. Maybe 5 years ago but not anymore. With proper head parking, more durable motors and better all round quality these drives will definitely stlll last at least 10 years before they die from sleeping.

3 140mm case fans? Are you running them full speed? Do you know the wattage? Are you using passive cooling for the CPU, or also running a fan on that? It's funny to see people try to build energy efficient systems and then not realize that cooling has a significant power impact.

All fans are set to the lowest fan speed that they can possibly stand while keeping the CPU less than 45 degrees. The case fans are running off the slowest option on the Define R4's fan controller.

I don't have any direct experience with the A6-5400K but I've heard griping that AMD's current power management is less aggressive than Intel's. Still, the most energy-efficient high performance systems we've been building here are Xeon E3-1230 based boxes that eat about 45W idle (no drives). A SoC-based NAS will take a few watts plus whatever the drives require, and an Atom-based NAS usually seems to average around 20-25W plus whatever the drives require (true of the name-brand NAS devices or if you roll your own Atom based FreeNAS). Intermediate platforms like the N36L/N40L/N54L are a little more than that, but really, hitting less than maybe 40W on a performance platform is going to be very difficult, and when you add drives to that, it pushes consumption up. Hoping for a 50W idle is probably unrealistic, but if you want to spin the drives down and can figure out what your drives need in order to make that happen, I'd expect 70-75W as a reasonable goal, and if there's anything else you can change in your system to reduce wattage, that'll help too.

When I manually spindown the drives (ataidle -S 10 ada0) I get 57w idle with the drives in standby. If encryption adds overhead to the power or stops the CPU from underclocking as quickly or other negative effects then there is a few watts saved there as well. Surely 55w is possible, and I thought maybe even 50w is possible.

 

[jgreco]

If you are dropping from 94W at system idle with disks spinning to 57W at system idle with disks sleeping, that's 37 watts difference, but the mfr's specs for your drives suggest 20 watts idle going down to 5 watts sleep, which is only 15 watts difference. The numbers appear to be inconsistent.

 

[jgreco]

If you're building a system from scratch, use laptop drives. These have much lower power usage, for obvious reasons. They cost more, though.

Careful with that line of logic though. If you're building a FreeNAS base system that's taking 40 watts and you get 4 10 watt 3.5" 4TB hard drives, that's 80 watts for 16TB total space, or 5 watts per TB. If you build it with 16 3 watt 2.5" 1TB drives, that's actually 88 watts for 16TB total space, but even in the most optimistic case you're going to need to add an extra controller (10 watts more), so you're up at 100 watts, or 25% more power for the "lower power" drive system, because their storage density is so much lower.

The calculations currently would appear to favor the 2TB 2.5" drives which seem to be running around 2 watts, but the price premium appears to be substantial.

 

[darkryoushii]

If you are dropping from 94W at system idle with disks spinning to 57W at system idle with disks sleeping, that's 37 watts difference, but the mfr's specs for your drives suggest 20 watts idle going down to 5 watts sleep, which is only 15 watts difference. The numbers appear to be inconsistent.

I re-run my tests, and actually found that setting the spindown to 5 minutes and the APM to level 128 (minimum with no spindown) then I run idle at 53w and under load at 75w - 80w. Keep in mind this is using a cheap $20 power meter I bought from my local electronics store, but that's what I'm reading. :\

 

[silvergoat]

Will a Kill-a-Watt work with an active PFC power supply? I think I remember reading somewhere that the numbers are inaccurate with some power supplies.

 

[jgreco]

Kill-a-Watt numbers are probably inaccurate if you need a lab quality accurate measurement. However, you're probably not interested in paying hundreds or thousands of dollars for a bench power meter, and if the thing says your power usage is reduced by 20%, that's probably close enough to accurate for the purposes people on this forum use them for.