[vineyardmike]

Can anyone explain a simple explanation of how chia works? I tried reading the White Paper ("green paper"), but its a little dense and math-y.

What is plotting? How does farming a plot work?

Edit: Why do people plot with SSDs? Is there a speed element to it?

[zamalek]

Farming: you hold massive (101GB) bingo cards of mathematical proofs on your hard drive. When you receive a challenge, you first do a pre-filter on your bingo cards to see which of them could contain a proof. Within that filtered list, you find the closest proofs. The peer with the closest proof available wins the challenge and [currently] gets 2XCH.

Plotting: you fill your free space with bingo cards. You need 300-400GB of free temporary space and TBs of writes in that space. This uses a lot of compute and eats SSDs for breakfast.

Once you have filled your space (assuming you don't go out and buy more disks), you can stop plotting. This is where the claimed "greenness" of Chia comes into play: you eventually stop using compute resources. Of course, there are whales who are adding drives to their servers daily.

[vineyardmike]

Can you explain the "bingo card" analogy in a bit more depth?

is this like pre-computing the values to a hash function against different inputs? Where is the write thrashing coming in? If you only need X hundred gigs, why do you write terabytes? Because you can't store intermediary values in RAM (since ram is expensive and small in comparison)?

I get how bitcoin style block-chains use hashing to secure the transaction chain, and the POW factors into hashing... how do responding to the challenges factor into a currency?

[zamalek]

> Where is the write thrashing coming in?

This happens during plotting (not farming). The plots are sorted, which is where the write thrashing happens. Once you've completely plotted a farm, it is never written to.

Everything else you said is basically spot-on.

[popotamonga]

I have 512GB Ram, cant it be used for plotting without ssd?

[mvanaltvorst]

You could. But someone else could add 8 SSD's for the same money and plot multiple times as fast as you, until the SSDs break.

[dstick]

And what happens if the SSDs break? You lose your plot? Has to be right - since it's proof of space? But what then is the use of burning through SSDs so fast?

[deepserket]

the SSDs are used just to create the plots, once the plots are created they can be moved into HDDs where they get farmed (checked to see if they can solve blocks for the blockchain)

[zamalek]

It can be, just set up a RAM disk and use that as your temporary drive. It's exactly what I'm doing to fill my tiny 12TB Raspberry Pi farm. Your plotting rate won't be competitive (plotting in parallel on NVMEs turns out to be much faster).

[lozaning]

Yeah, I plot an a DL580 with 512GB. Other than not having to worry about write durability there's not much gained though. GB per GB my ram drive is faster, but i can plot in parallel on my NVME drive because it's considerably larger.

[VectorLock]

You could do 1 plot at a time.

[Vespasian]

Is there an expected limit of storage where your average drive would die of old age before it yields enough income?

[zamalek]

No. Once you are farming, the drive is exclusively read. Of course there is bitrot/cosmic rays flipping bits, effectively rendering some portions of the plot corrupt.

[duskwuff]

In short:

Plotting is an compute- and IO-intensive process which eventually generates a ~100 GB block of data after a large number of read/write operations to that data. This is the part of the process that involves (and destroys) SSDs.

Farming is a slower process which is performed on those blocks of data, which involves occasional read operations on small sections of that data. The data blocks are typically transferred to slower bulk storage (like hard disks) for this.

[fbcpck]

If you'd like it in video format, [this one][1] explains it in layman terms quite well and covers possible ways to mitigate the ssd endurance problem, as well as expected profit rates with the right "commodity" hardwares.

[1]: https://www.youtube.com/watch?v=twwyBdsRYL4