Does anyone know how many watts a single 3.5" HDD consumes when hosting only a Storj node? I’m wondering if I can expect 10W per drive or 5W per drive.
I’m also unsure if SAS drives consume more power than SATA drives.
Any information would be really appreciated.
You should be able to find the specs for almost any HDD these days. Standby around 1w, idle 3-4w and and active 6-7w are common for modern models (example).
Because a 4TB HDD and a 24TB can be using similar power… there’s an obvious bias towards using larger drives for SNOs that may be paying that power bill for years. So many people look at it as watts-per-TB of their entire setup: where 1-watt-per-TB would be ideal.
Yup. the disk is in constant action so i count average like not less than 8,5watt per disk per h for my 16TB’s SATA. Some disks constructions are more efficient and over all better than others, for example 10TB ultrastars vs 16TB ultrastars, 16TB’s are better made here, but its a difference of a 1 or 1,5watt per disk, but still.
Electrically they’re the same. Firmware or interface changes should have negligible impact on power usage. This is a sample data sheet: if you compare models that differ only in interface (e.g., ST16000NM002G and ST16000NM001G), you can see the specification of power usage is almost exactly the same.
As such, what matters is mostly the amount of time your drive idles. With a Storj node actively accepting uploads, the drive will probably never really idle. But a full node will get less and less activity with time.
even if a storj drive was full, there would still be multiple reads per second so the drive would never spin down, even if the actuator head isn’t banging around for a few milliseconds So power usage, I suspect, is always going to be closer to the “full power” range than not.
Maybe except solutions with the SSD cache for at least metadata.
I had nodes full for several months having disk accesses mostly due to audit checks. They’re rare. It’s not going to be that easy anymore with a lot bigger data rotation, but I still find this scenario plausible.
Hm, this is not what I see for my nodes. GET_AUDIT requests are negligible compared to GET and GET_REPAIR. 
Wait a year with no ingress and most of these GETs and GET_REPAIRs will disappear 
No ingress and still full for several month… How to get this?
This used to be easy with static test data. Now it’s more difficult, which is why I wrote: