So what is your watt per TB figure?

We have had a lot of posts and photos about people’s storj rigs, but in today’s climate, the energy costs of a system is becoming far more important.

I would say that the best numbers would be as I detail below for dedicated systems. For someone that is operating a shared storage system such as a NAS such costing does not make sense, but someone may be able to work out their extra costs if they have an environment where the disks would spin down if there was not a constant load from storj processing.

System - Pi4, external 12TB drive and USB hub.

Total system power draw - 16W
power draw per TB stored (6.17TB) - 2.6W
power draw for a full node (12TB) per TB - 1.3W

My interest comes from the long term running costs of a node as energy prices keep rising. A 16W power draw means a yearly total draw of 140kWh. Last year this cost me around £13, this year £20 and next year when I renew my contract it could be £50 if not more.

{Modified as I typed TW rather TB in key places}

2 Likes

System - Supermicro server, 2x Xeon X5687, 108GB RAM, 4U with 24 hard drive slots and two power supplies, 12 7200RPM drives (6x4TB + 6x6TB) used in the pool that stores Storj data (and some other stuff).

Total system power draw (currently) - 362W (it jumps around somewhat)
power draw per TB stored (19TB) - 19W
power draw for full node assuming I do not add more drives (~32TB) - 11.3W

I use the server for other stuff, not just Storj (though if I did not run a node, I probably would do some things differently)

System Tyan server, 2x Xeon L5640, 114GB RAM, 2U 12x 3.5" BAYS Single PSU.
disk shelf with 24 Bays… 30 disk’s in the system, but many not used for Storj and server being used for other purposes also.

Total system max draw…less than 400 to 500watts
unsure server itself is between 220 and 300 depending on cpu load.
the disk shelf is between 30-40 idle and empty, really should get that measured with disks.
but i would expect it to be less than 24 x 6 watts (which is the hdd max draw(according to spec) so 140 + idle of the shelf.
(might update this when i do some measurements in the near future.)

total usable storage 86TB / 400 to 500 watts
Watts pr TB capacity = 4.6 to 5.8 (after raid redundancy and disk partitioning)
might be lower will need to measure it exactly.

been trying to improve my TB to wattage ratio, because my power is pretty expensive here.
also thought about maybe setting up some solar cells for it.

A HP Microserver of an older generation with four drives totaling 14TB, most (but not all) for Storj. 50W total measured at the wall, so about 3.5W/TB. Thinking of getting one larger drive next Black Friday. I could theoretically go for a set of 4×18TB with almost the same power draw, but I just don’t need that much space, nor Storj would be able to fill it.

Maybe spin down is not available but I have read that some drives have a quiet mode which is short of spin down but saves power

True, many drives will park their r/w heads if not dealing with requests and sit in an Idle state/mode, but the real power saving comes from any situation where they can stop the spinning of the drive.

This is a nice write up on the subject from the days when disk capacity was in the GB range, rather than TB.

 http://ixbtlabs.com/articles2/storage/hddpower.html

However, hard drives have limited number of start/stop cycles or park cycles. If they do so frequently, then they will fail sooner. I usually set it so my hard drives never stop or park their heads.

For example, on my file server (not used for Storj):

  4 Start_Stop_Count        0x0032   100   100   000    Old_age   Always       -       38
  5 Reallocated_Sector_Ct   0x0033   200   200   140    Pre-fail  Always       -       0
  7 Seek_Error_Rate         0x002e   200   200   000    Old_age   Always       -       0
  9 Power_On_Hours          0x0032   017   017   000    Old_age   Always       -       60611
...
 12 Power_Cycle_Count       0x0032   100   100   000    Old_age   Always       -       38
192 Power-Off_Retract_Count 0x0032   200   200   000    Old_age   Always       -       20
193 Load_Cycle_Count        0x0032   197   197   000    Old_age   Always       -       9400

9400 load cycles in 60k hours. It should have been lower, but it was a while before I found out that the drives were doing that and decided to stop them.

It depends on how you count. For me it’s really not a factor, I use only systems that were already online 24/7. I did actually test the difference between running the nodes and not running them and that added up to about 5W difference.

So 5W/19TB=0.26W per TB

Total system power is about 105W. So if you count it all it would be 5.5W per TB. But to me that doesn’t make sense, since I would be paying for 100W even without Storj running.

No, the marginal cost is the right cost to use in such a situation - the extra cost to perform an extra task, so for you, it does not matter how much your platform costs in terms of energy usage for it to perform its main/primary task, just how much extra energy is used to also do Storj based tasks.

This is one of the reasons why I said we should just compare the energy costs of our systems as other marginal costs - such as disk space or system wear and tear are far harder to generalize.

RPi 4B with 4 disks (3x2.5" + 1x3.5").
Dedicated to Storj. I use it for small tasks sometimes, but it’s essentially for Storj.

Total draw: around 30w.
Max capacity: 16TB (1.88w/TB)
Currently storing: 6.5TB (4.62w/TB)


I can’t see how this would be possible when running Storj nodes.
Disks never rest, they never stay more than 10s without having to query or store a piece of data. And that’s during quiet moments!

Are disks still capable of reading and writing data is that mode?

Yes of course the drive can read and write!
The thing I don’t know is which mode uses the lower power. Some people say that performance mode uses less power than quiet mode

Ah alright, then I guess this would be a cool option to reduce power usage (if quiet mode reduces power usage…). Didn’t know such a mode existed on some drives.

Wait what? That would be so counter intuitive! :open_mouth:
If that’s the case, then this mode wouldn’t be useful for running Storj I guess.

Would you happen to have a source for that?

i assume it’s this andrew is talking about.
not sure if it actually saves power, none of my hdd’s supports it but they are kinda old.

Mulipurpose QNAP 16 bay NAS - 100w (online anyway)
~70TB available to use
Storj is using about 18TB, I expand as data scales up.

Multipurpose Proxmox Hypervisor with docker container - 50w (online anyway)

So, i have about 18TB /150w = 8w/TB
Or, really only extra hard drives are my only overhead ( 4 drives @ 9w/ea)
18TB/36w = 2w/TB