Saturday, June 12, 2021

I wish Chia farming could work with offline drives (or maybe it can?)

I have lots of smaller (mostly 2TB and 4TB) drives that would be great for storing some plots on. I also have a 15-bay disk shelf. If I stuff it full of all the drives I have available I can add close to 50 more TB to my plots.

Here's the thing. When the shelf is powered on, with all the drives spinning, it sucks quite a bit of power. Not an unmanageable amount to be fair, but easily around 200W idling. The shelf's controllers and fans alone use around 60w constantly, and the drives use maybe 8-10w each just idling.

What I wish (and wonder) is that Chia could somehow keep a catalog of what plots I have, and just enough data to know if I have a qualifying plot. That way, my entire disk shelf could be on a smart switch, and spend the vast majority of its time powered down. In the event that I have a candidate plot, a signal could be sent to the smart outlet which would power up the shelf. The same script would monitor for the shelf to appear on the SAS interface, and once it does, the drive(s) would be mounted and whatever else needs to be accessed would be accessed at that point. If I for some reason had deleted the plot, that would of course be discovered and I'd lose out.

One of Chia's goals is saving energy by using storage instead of compute power. While I'm still using a lot less power than anyone would use for ethereum or bitcoin mining, it still seems like a lot of wear and tear on drives that are just spinning forever, waiting for a rare event.

For people doing small single-drive or low-number-of-drives mining on e.g. a RasPi, you could just use a drive that spins itself down. But most enterprise drives don't even bother to spin down on their own, and even if I do spin them all down manually (assuming they can do that), the shelf is still using constant energy.

Anyone else considered this thought? Or am I just completely misunderstanding how Chia farming works? (I am quite new to this after all, so apologies.)


No comments:

Post a Comment