Homelab - Datacenter style mining rig feasibility

So guys, I’m not necessarily new to Chia, I mined quite a bit when it was just a little baby, however I’m thinking of trying to mine in order to justify buying more storage, I’ve got a full on homelab in another country that I use for all of my production stuff and after getting a new job I want to effectively replicate that in my home country, however I’ve got a pretty significant issue, I’ve got really bad internet here, however I’m getting good pay, so I was thinking rather than trying to sync > 150TB Of data over a 70Mbps network I’d just mine Chia until I’ve got good internet.

Since I’m spending a bunch of money on stuff anyway, two DS4246 Shelves with an R730 With dual 2699 V4’s loaded with 336TB of drives in one and hoping to fill the second one up over time from Chia profits, a 2TB NVMe in the R730 just for plotting and then doing CPU Mining on one full CPU, is it feasible for me to in say, a year’s time make 7000 USD to fill out the second drive shelf, because that should be when I get new internet and if I’ve got an extra ~300TB of drives for “free” then I’d be pretty damn happy, or I could just keep mining on the second shelf and utilize the first for what I actually want to use it for.

The big worry for me is drive reliability, I’m looking at Enterprise Hitachi 14TB Drives which are rated at 550TB/Year, I’d hit that in… 26 Days if the drives are running at full tilt where they can do ~240MB/s, how long are people finding their helium drives lasting? Are hard drive companies caring about warranty stats anymore? I was thinking of just flashing garbage firmware on them if they die so they can’t really check for how much data has been written.

Many thanks

Considering the extremely low usage of HDDs by Chia and the conversations on this Forum, HDD life is probably the least of your concerns.

Farming many Chia plots with a bad internet connection will almost certainly be your biggest problem. With current Chia software issues, downloading the DBs with a bad connection can also be slow, difficult, and sometimes a fail.

That’s interesting, when I say a bad connection I mean relative to the amount of data I currently deal with, I’ve got a 72/14 Or so connection, thanks for the insight on what is going to be more important, what are the typical loads looking like right now on the drives / should I just be looking that up instead of asking?

While farming, they will be more aroundt 2kb/s…

It is such a non-issue that it is difficult to research. Most threads end up talking about SSDs and NVMes after dismissing any HDD concerns brought up.

When farming, Chia only checks plots that pass the filter of 1/256. Chia farming HDDs are only briefly accessed when a plot on a drive passes the filter. Chia HDDs should never sleep and do not suffer sleep related on/off stress. Chia farming is one of the least intensive uses you can put a hard drive to. The only real HDD use is when plotting to it, and even this is one time (hopefully) and only writes the entire drive once … not much in any drive’s usage.

Even the concerns about eating NVMes (and SSDs to a lesser extent) have proven to be highly overblown. There is a long-running thread here at the Chia Forum where I think no-one has yet reported killing their NVMe. It appears that heavy over-usage (Chia plotting) may (not sure) effect the NVMe’s ability to hold data after power down. Even if true, this has no effect on Chia plotting as you never power down while plotting.
The NVMes may not be good for anything else after you have written three or four times their rated TBW but they are hard to kill if kept at their Chia task.

I misunderstood.

You should have no bandwidth issues adding Chia to your setup.

In terms of HDDs, while not ignoring quality, price per TB would be my primary concern.

There are a few threads here where peeps point out some very good deals. When I buy more drives it will probably be from hearing about it here or on amazon.

Personally I prefer the WDs over the Seagates and do not have much experience with Hitachi.

As I remember someone else saying, most bad HDDs will fail on day one or close to it, so buying a drive with easy/good warrantee with easy return/replacement would also be important to me.

Good Luck and Be Well! :grinning:

Thanks for the detailed response!

I’m glad to hear all of this, as the cheap part of my system is the one part that has the most potential for failure (NVMe Drives) but even then, I’m used to them being hit pretty hard and this is separate from my boot drive, I’ve got a drive with 2PB Written to it from other stuff and it’s just… fine.

I’ve worked in datacenters and at an MSP I’ve found that drive quality really just doesn’t matter after a certain point, conditions do far more, at my previous employer we would quite literally only buy refurb seagate drives because they lasted on average one day less, big deal when it’s saving you $50 per drive and you’re buying 10 at a time

If it loses the plotting data what would even happen? Just need to start plotting again while all previous plots on the drives would be fine?

I’ve got a lot of 980 Pros as spares if anything goes wrong :smiley: again, thanks!

1 Like

I have a 2TB 970 plus which has proved brilliant so far. I have heard nothing but good about any of the Samsung NVMes and appreciate the excellent Samsung Magician software.

You have the right NVMes imop! :+1:

Yes, but even better, you see most drives failing long before they FAIL and would be able to save/transfer most of the plots before they were damaged. :sunglasses:

That’s certainly something that will make life easier, I doubt they’ll get that bad over the course of a year even.

Thanks for the help, can’t wait to spend far too much money on my “Chia Farm” :smiley:

1 Like