Proof of Space with RAM

[pwn8]

is it possible to make a proof of space that is stored entirely in RAM?
 in my opinion proof of space has failed because storing in hard drives is not practical and ends the useful life of the hard drive.
If there is a way to have the voting power based on ram memory, in my opinion,
it would be fairer than proof of work.
i am thinking of just storing hashes in memory and the winner is the hash that is closest to the hash of the previous block.
an attacker can just recalculate the hashes like in conventional pow but in the long run storing the hashes will give an advantage.
if the miner doesn't have a valid hash he can start calculating other hashes with the CPU.

[1714514917]

1714514917

[1714514917]

1714514917

[1714514917]

1714514917

[1714514917]

1714514917

[digaran]

What? 🤔😂. Where should we store the blockchain data then, on RAM? Hashing needs computational power >CPU, People need to verify data on their own>Blockchain data on storage device. Ram is just a temporary and fast storage device.

[decodx]

I think it is theoretically possible to create a proof of space algorithm that utilizes RAM instead of hard drives for storage. However, there are several challenges that would need to be overcome to make this approach practical and effective. The biggest challenge in my opinion is that using RAM for storage could make it easier for attackers to create fake proofs of space, as RAM is more easily manipulated than hard drive storage. RAM is more volatile and can be more easily manipulated than hard drive storage, which could make it easier for attackers to create fake proofs of space or manipulate the consensus process in other ways.

[logfiles]

Maybe you should know the difference from ROM, Long term storage and RAM before you even think about your proof of space  

Why do you think so many poSpace projects are using HDDs, SSDs and not RAM?
It's not a mistake. RAM is temporary. How are you going to keep your blockchain up and going if the data stored in the RAM is just going to be temporary?

[tromp]

is it possible to make a proof of space that is stored entirely in RAM?
 in my opinion proof of space has failed because storing in hard drives is not practical and ends the useful life of the hard drive.
If there is a way to have the voting power based on ram memory, in my opinion,
it would be fairer than proof of work.
i am thinking of just storing hashes in memory and the winner is the hash that is closest to the hash of the previous block.
an attacker can just recalculate the hashes like in conventional pow but in the long run storing the hashes will give an advantage.
if the miner doesn't have a valid hash he can start calculating other hashes with the CPU.

Proof of space is inferior to proof of work in that, in order to prevent grinding,
it doesn't directly commit to the transactions in a block [1].
It only commits to a miner private key.
The miner then separately provides a signature to commit to the transactions.
That means that a group of recent miners can collude to rewrite recent transaction history.

A memory bound proof of work, like Cuckoo Cycle [2], seems like a better form of Proof of RAM...

[1] https://docs.chia.net/consensus-foliage/

[2] https://github.com/tromp/cuckoo

[DaveF]

ram memory is much more accessible and "equal" than cpu and then asics.

So then we could call it the Amazon coin. Since they have datacenters filled with servers just sitting there waiting for there to be a need for them during periods of high use.
They could just kick them on to mine your PoS RAM coin.

There is no need to change the way things work now. You buy a miner and you mine. If you can't buy a miner you buy coins. If you change the algo from SHA256 to something else it's just a matter of time  until someone starts building purpose built miners and the people who can afford them get them and those that can't are back where they started. Possibly even worse off if they invested in something that is now obsolete.

-Dave

[decodx]

ram memory is much more accessible and "equal" than cpu and then asics.

How precisely is RAM more accessible or "equal" than the CPU? RAM is just another chip, and we are still reliant on manufacturers and face constraints with market supply. Unless there's a method of producing your own, the situation remains the same.

[tromp]

As I understood more RAM does not increase the chance of winning in cuckoo cycle is it possible to adapt this algorithm to do this? to make a proof of RAM algorithm?Maybe run several times in parallel?

Cuckoo cycle performance depends on several things.

1) memory transaction speed, i.e. how many random bits you can read/write per second.

2) you need to compute a simple siphash24 hash function per memory access

3) You also need a minimum amount of memory, e.g. 512MB of SRAM and 512MB DRAM for Cuckatoo32.

You need to balance 1) and 2) so that neither bottlenecks the other.

ram memory is much more accessible and "equal" than cpu and then asics.

RAM by itself it useless without the accompanying computing cores to use it.
That is, assuming you want to proof of work to actually commit to the contents of the block,
so it has to be computed from scratch all the time.

We can use cuckoo cycle but instead of increasing the difficulty like in hashcash we increase the amount of memory required. Being asics resistant by design.

That doesn't work too well, since increasing memory also increases the time per solution attempt, and you need to keep that time a tiny fraction of the block interval to keep mining behaving like a random lottery.

[decodx]

cpu can have multiple cores. ram memory can have variable speeds but it is very little difference. high end ram memory does not have as much advantage compared to a high end cpu.

It can make a big difference depending on the amount of RAM available. For instance, if you have a desktop computer with 8GB of RAM and I have a server with 4x32GB of RAM, are we on an equal footing? If I manage a server farm, the difference could be even bigger. And don't even get me started on the sheer superiority of the data centers and cloud providers owned by big corporations.

if a proof of ram coin enters the market it is only a matter of time for decentralization. we could say the same with bitcoin in 2009 because amazon had big servers.

Are you seriously considering repeating 2009, just because you missed out on the chance to mine Bitcoin when it was dirt cheap? Have you even thought about the future of such a coin? What if someone starts complaining about RAM coin in a decade or so because the mining process is not decentralized enough? What should we do then?

[mikeywith]

if a proof of ram coin enters the market it is only a matter of time for decentralization. we could say the same with bitcoin in 2009 because amazon had big servers.

It would be just a matter of time before someone starts making larger rams that are capable of doing the job better than the average ram, and even if that fails, the large players will end up owning enough RAM that makes everyone else's ram pretty useless, ASIC miners came as a natural result of people's desire to make more money more efficiently, it doesn't matter how you want to start the protocol, RAM, CPUs, GPUs or even light bulbs, the system will evolve and we would still end up with something like ASIC miners.

[LoyceV]

It would be just a matter of time before someone starts making larger rams that are capable of doing the job better than the average ram

Or just write it to disk That's how "proof of space" works: you don't need to access the data frequently, so you don't need to have it in RAM.

There's an alternative that exists already: ACIS-resistant protocols that require a large amount of RAM to mine. It still needs processing power, but you can't really avoid that if you don't want the content of the RAM to be stored on disk.

[mikeywith]

There's an alternative that exists already: ACIS-resistant protocols that require a large amount of RAM to mine. It still needs processing power, but you can't really avoid that if you don't want the content of the RAM to be stored on disk.

You are correct Loyce, the point I was trying to make is that no matter what protocol you use in order to achieve any sort of decentralization is always going to end up in the same way,  ACIS-resistant protocols, a recent example of that is Chia coin and its protocol which is based on both proof of space and time, all you need is a hard drive to store the plots, and a processor to make those plots did not take long for someone to figure out a way of writing that space with RAM in a faster and cheaper way.

The argument people had of how that protocol was more decentralized than BTC's POW is that fact that people had better access to HDDs than to ASICs, it took only a few months for that theory to go out the window, videos of large farms with thousands of HDDs were floating around, in just a few months every average Joe with a few 10TBs HDDs was not able to make any profit.

It doesn't matter what protocol you use, the end result is going to be pretty much the same, people that know how to make money will always find ways to make more money, even when you give the average Joe a chance to make money he will miss it, so thinking that ASIC miners are the problem and by using something else absolute decentralization is going to be achieved is daydreaming.

[dansus021]

I think it possible but will be not efficient since the price of ram is very high and If we mine on RAM requiring a dozen operation of write and read continuously.

Why there is still Proof of Storage it because people save the data on it and the price of storage per Gb is relatively cheap and if we using mechanical hardrive it will much cheaper But RAM is expensive right?

[LoyceV]

I think it possible but will be not efficient since the price of ram is very high

Bitcoin miners aren't cheap either. That literally doesn't matter, since all miners have to deal with it. If it's cheaper, they buy more.

[PawGo]

I think it possible but will be not efficient since the price of ram is very high

Bitcoin miners aren't cheap either. That literally doesn't matter, since all miners have to deal with it. If it's cheaper, they buy more.

ASICs are not cheap because they make a good profit. And in long-term I think hardware costs become a secondary problem, as electricity/cooling/hosting etc becomes more important.
But it is interesting to have a wide view, how electricity and hardware costs may impact profitability and the whole network power (and difficulty adjustment).

[Skybuck]

Perhaps some "Proof of Links" or "Proof of Access Speed" could be created.

For example a Linked List is harder to traverse than an array.

However there is a list ranking algorithm which can run on multiple threads/cores/processors to process the list faster.

Alternatively some kind of network of links could be created in RAM, basically describing some kind of path through the RAM, through cells.

It could be 1D, 2D, 3D or N-D dimensional structure and links.

Some paths/sub paths/links do exist while others don't.

Thus it takes some time to find the way throught the memory. Basically like a MAZE.

Perhaps "proof of shortest path" could be interesting.

A challenge for computers to find the shortest path through a MAZE in memory !

(BTW I think there might already exist some kind of proof of ram or something in some cryptocurrency)

[tromp]

Perhaps some "Proof of Links" or "Proof of Access Speed" could be created.
For example a Linked List is harder to traverse than an array.
However there is a list ranking algorithm which can run on multiple threads/cores/processors to process the list faster.
Alternatively some kind of network of links could be created in RAM, basically describing some kind of path through the RAM, through cells.
It could be 1D, 2D, 3D or N-D dimensional structure and links.
Some paths/sub paths/links do exist while others don't.
Thus it takes some time to find the way throught the memory. Basically like a MAZE.

How about finding a fixed length path through a random maze-like network of billions of nodes?

That gets back to where it started?

Which requires following billions of links through cells in RAM?

(BTW I think there might already exist some kind of proof of ram or something in some cryptocurrency)

Yes. It's called Cuckoo Cycle, as I mentioned before: https://github.com/tromp/cuckoo

[dansus021]

Bitcoin miners aren't cheap either. That literally doesn't matter, since all miners have to deal with it. If it's cheaper, they buy more.

Yes but Bitcoin miner is isolated like to crypto miners only.

If Proof of Space is really made a different chain and explodes it will be more like GPU or at least it will be much worse than GPU since RAM is essential to a computer. It needs RAM to work, and GPU on other hand computer can work without GPU.

But the real question is whether we can make a chain based on Proof of Space and replace the great traditional mining ex. Bitcoin Mining

[PawGo]

Even with current Proof of Space/PoS which use storage drive, it only cause quick shortage since PoS miner realize they need SSD with very good quality/durability. So i expect if coin which use algorithm which stress the RAM become popular, people would buy high-end or pricey RAM which won't lead to shortage among average computer user.

Probably you came to correct conclusion, even today when we talk about -let's say- ECC, we may have consumer-targeted RAM and professionals-targeted, server RAM, where ECC is required and is more important than a "speed". Like many other computer parts, also memory could be 'tuned', we may increase freq - but it is not the way how serious servers work, as they expect absolute confidence of results.

[dansus021]

But have you every try using computer without any GPU? Using CPU for 2D graphic and playing video would lead to poor performance (30 FPS or lower) even when you have fast CPU. I already experience that since i tried using Windows VM without any GPU acceleration.

Even with current Proof of Space/PoS which use storage drive, it only cause quick shortage since PoS miner realize they need SSD with very good quality/durability. So i expect if coin which use algorithm which stress the RAM become popular, people would buy high-end or pricey RAM which won't lead to shortage among average computer user.

Yes most of the laptops nowadays use integrated Graphic Units. And a Powerful CPU like Ryzen APU can run some of Triple-A games.


In my opinion PoS is not yet very popular like GPU mining and I think Disk is cheap to make per Gb rather than RAM. But the RAM is different "RAM Is Expensive To Make
Extremely complex equipment is required to fabricate the chips and, because impurities in or on the silicon can cause defects" - https://www.makeuseof.com/tag/how-is-ram-made-and-why-does-the-price-fluctuate/

and manufactured need a couple of years to make a better efficient RAM ex. DDR3 2007 DDR4 2014 so when there is mining proof of space with RAM it made shortage on consumer or enterprise level just IMO