Is farming viable?


#21

But you are also competing with people with little or no infrastructure set up or running costs. They are likely using unmetered broadband, with hardware bought to do something else, sipping at the electricity supply.

Data centres will almost certainly provide the highest performance, but that doesn’t mean home vaults aren’t useful. They will take the slack when needed and will keep the whole network honest.


#22

Thats my point, big machine in datacenter provides best performance and because (if I understand it corectly) the farming reward is based on latency, how do we make sure there is something left for the smaller slower home vaults?


#23

10% of internet users: about 400 million (end of 2017)

20% of the vaults owned by home users: 20% of the users (whales) own 80% of the storage space (Pareto principle) so 20% of the storage is owned by 80% of users (regular users)

1% of the coins recycled a day: 2^32 * 0.01 / 86400, so abut 500 Safecoins redistributed each second

If there are 400 million home users who own 1 vault, there will be about 5 times as many vaults, so about 2 billion. If each stores 1TB and the cost of 1TB is $25 then the total storage is about $50 billion.

As rent is always connected to real estate value, renting that storage shouldn’t cost much more than how much the storage cost, so the market cap can’t be much higher than $50 billion ($12.50/coin) when the network is an everyday thing, after the hype cycles are done.

This calculation isn’t affected by cheaper storage because we always use as much as we can pay for. If the Safe Network happened today, this is the approximate maximum storage a vault would provide.

Wrong word, sorry. Recycled. I used it as such in the calculations.

Which is a legitimate method.

It’s important to place these numbers in a rough ballpark to see if Safecoin and farming can work when the network gains worldwide adoption.

If my node sends millions of transactions, that’s spamming the network. Isn’t already something in place to prevent that?


#24

Yes. However, as the network fills up with data, more and more coins are already allocated and a less and less number of the coins can be farmed. It’s a dynamic balance.


#25

Your own doubts about things will make this principle not applicable to SAFE.

The extra costs for running data centre vaults over home near zero cost, will work against the 80%/20% since the data centre vaults will not be as profitable as home vaults.

Home vaults will be ever so easy to run, the economics of scale doesn’t work here, unlike what the principle is based upon.

The storage will be greater than what people want to store. Since we do not have anything like SAFE on the current internet then you cannot estimate total storage that way. I expect that as soon as people can store securely their private data then they will be uploading a lot more than has been seen ever.

Even if there is a high %age of vaults as datacentre vaults then storage for them is a lot higher than for home users. So 25$ per TB for data centre permanent storage isn’t realistic at the moment, since datacentre costs has to include VM instance rental, data storage rental etc.

Of course trying to work out a market cap for a utility coin that has recycling is a lot more complex then total storage multiplied by a $/TB The coin will have a lot of utility for people paying others for goods etc. So there is also value in the coin itself.

But still the figures are just a guess. This was my point

If they are valid then no its not spam. Spam will be trying to transact one coin address to multiple IDs In any case the wallet can mitigate that by various methods.

As it fills then people are encouraged to add more storage by the dynamic farming rate

But you are forgetting that the data has to be uploaded in order to fill up the network. Also the cost to PUT will be much more the reward for GETs.

Obviously the algorithm has to be tuned otherwise you can easily get the situation where too many coins are farmed for the coins recycled.


#26

Yes, I know: “this time it’s different”.

The figures are close to the theoretically possible values. The results may be 10x less or 10x more, but not 1000000x less or 1000000x more. Therefore, it’s a useful estimate.

At any given moment in time, the number of coins burnt (paid) and reallocated will be very close to each other. That’s the only thing I needed for the calculation. If 2 billion vaults compete for 500 farmed coins/second (if 1% of all coins are recycled each day, which is a high estimate!), then it takes a long time before any one vault can farm a coin. And that is a problem.


#27

Again, it can’t be very far from the storage value. It can be 2x as much, maybe even 10x as much. It can’t be 10000x as much.

This coin is not very useful for everyday spending. I can send arbitrary amounts with Bitcoin and friends but Safecoin comes in one size only. I saw your suggestion about smaller coins. Great, small values are solved. But sending a million coins for large transactions? Do you seriously believe that’s how it should work? I looked through a few of the division threads as @jlpell suggested and saw a few different versions of coin denominations. Those seem more realistic.


#28

I think some free SAFE quota will be needed to get enough users. And for sufficient network growth. I’m thinking about the majority of users, not the early adopters. Not even farming will be viable if there is insufficient end user interest, because then too few GETs are generated to make the network grow and to reward farmers with safecoins.


#29

Why?

Coins are only paid out as GETs occur. You assume that the number of rewards for GETs match coins paid & destroyed. That is not an assumption that can be made. One simple reason is that the algorithm will need to be tuned by necessity to mitigate such a scenario occurring.

Your calculations seem to be based on farmers being paid for amount of storage supplied. Whereas it is based upon the successful GETs they supply. And the number of GETs in a period is very variable.

And that is why coin division is discussed elsewhere a lot, and is going to be introduced.

But they aren’t really since the transaction load increases a lot. As you pointed out transacting large numbers of coins is not good.

It has been suggested that this may happen. Most likely the release candidate betas will have a PUT balance allowance to get over the chicken & egg problem. But as soon as you give free PUTs in any system then storage spamming will destroy the system by filling it up as we have already seen in a previous test network.


#30

Because anything else means the network is not functioning correctly.

There are 4 billion coins. A fraction of those are free to farm and the rest belong to wallets. While some coins will be used to buy PUTs and others will be farmed for GETs the ratio of those two fractions needs to remain relatively stable over short periods such as days or weeks or else the network would be highly unstable.

You sure don’t expect that one day 45% of the coins are farmed, the next 87%, and then 31% the 3rd day, right?

What do you mean? If there are larger than 1 value coins then the transaction load would decrease not increase. I could pay 10,000 Safecoins by sending a single 10,000 coin instead of 10,000 x 1 coin.


#31

I do not see how you can say this. It does not match anyones model of how things will work.

What do you understand farming is?

I was thinking you were talking of the SAFEcents and SAFEmillicents denomination idea.


#32

It’s basic necessity. A system that can’t maintain balance can’t function.

There must be mechanisms that slow down farming as the network starts running out of unallocated coins. At the same time, coins must become more expensive to discourage using up space as it becomes scarce. These two have to be able to create a balance at any level of demand and supply, or else the network can not function.

As long as demand and supply doesn’t change drastically, the ratio of allocated vs unallocated coins will also not change. Users will farm and pay about the same number of coins.

That wouldn’t solve transferring larger values.


#33

Yes, that’s tricky. In order for farming to be viable one necessary component is to have enough users using the system. And with free quota, there has to be control over user registrations to prevent people from creating lots of accounts. Many people will need more than one account, but in those cases it’s fine to have only one account per user with free quota, and there can even be a cost for creating more than one account per user.


#34

Well I still don’t hold to the idea that just because people upload a lot and spend X coins that there will be X coins farmed immediately to match that.

Have you investigated the scarcity part of the farming algorithm? As more coins exist then less of the successful farming attempts will result in a coin rewarded.

Now I am thinking by your assertions that you might have some misunderstanding of what farming actually is. Farming is retrieving data and rewards are only given for successful retrieval of chunks.


#35

That is why the last release candidate beta will turn into the live system. This way there will be viable data already when the system is live. People will never know which release candidate beta will become live so they will be uploading good data just in case it does.

And the network has no way to know if you do. Its an anonymous system designed to not know these things


#36

I forgot, the SAFE network is meant to work without the web. I was thinking about paying for the SAFE accounts via some website or exchange, but then SAFE becomes dependent on the web, not good.

Another idea is to make farming super easy to get started with, even on mobile devices. So then users register SAFE accounts for free, but they need to start farming, buy or receive safecoins from someone else before they can store their own data.


#37

I couldn’t respond earlier due to forum restrictions.

It’s basic control theory. The system need to ensure that its parameters stay within sensible ranges. Every living thing works like that too.

The mechanics are irrelevant so I might as well completely misunderstand farming and yet the conclusion would stay the same. If the network can’t ensure dynamic balance between spending and farming under all circumstances then it will run out of either storage or coins and it will fail. Therefore, I assumed the network does maintain a balance (or else there’s no working network to talk about), and that implies the ratio of farmed and unfarmed coins is relatively stable. In other words, about the same number of coins are spent as the number of coins farmed.

Note: I know there is a mechanism to slow down farming (the farming divisor). “There must be such a mechanism” was part of the argument and it wasn’t meant to imply such mechanism wouldn’t already exist as part of the design.

Ironically, I don’t hold to that idea either.

It’s not “just because” and not “immediately to match that” that those two numbers will be close. Two things can co-occur not only if one causes the other but also if a common thing causes them both.

In this case, however, the balance would just emerge where the two curves that govern PUT costs and farming rewards intersect. As long as the shapes of those two curves are fine, spending and farming will always be slowly nudged towards where they are at equilibrium. This point will keep changing of course (probably depending on the demand/supply ratio, the real-life value of the coin, and other things) so it will never be “reached” just continuously approximated. However, it will be relatively stable over a few days or weeks.

But this was just one of the assumptions for my calculations. I thought it would be immediately clear to anyone why it is necessarily correct. The network couldn’t stay stable otherwise, it’s that simple. I’m really baffled at why anyone would think there’s something deeply questionable or radical about such a basic and obvious assumption.


#38

In the beginning farming will likely be done by tech savvy people using PCs. Later on however it would be great to be able to farm with mobile devices. I haven’t checked but I guess that the majority of user growth in the world is now in the mobile space.

Farming is viable in different layers, all the way from archive nodes to mobile nodes with tremendous churn. I was thinking of how in a computer, the memory is accessed in several layers, from registers to fast cache to slower cash to RAM and to the hard drive. Similarly, in order to make farming viable for a broad spectrum of use cases the mobile devices can serve cashed chunks of data, and have a different category than regular farms because otherwise their ranking will drop too quickly since the mobile devices hardly can be up and running 24/7.


#39

You are assuming @JoeSmithJr that the price of purchasing PUTs will always follow the value of the coin. That’s not necessarily true. The coin is independent of the PUT cost, and could “moon” in price without the price of PUTs changing for a time. After the network realizes PUTs aren’t being purchased, it raises rewards and slowly lowers its PUT price until it gets back in “equilibrium” with the price people are willing to pay for PUTs versus the amount of free resources.

This is where big farmers who can scale quickly will come in handy. They can spin up new farms quickly to take advantage of these lulls in available storage. Every day Joe’s (no offense), by word of mouth, will also hear of these big rewards and start a vault. The big farmers who spin up to receive the big rewards spin down as new people create vaults because it becomes less cost effective.

Storage may become so cheap with this method, that a TB could cost $1 even if it’s $25 to purchase a TB of physical storage space. The cost of storage in the network is not dictated by real world prices but by availability of resources. Also, your cost are way off, anyway, because you have to remember there are 8 copies. So if you are estimating a TB to cost $25, it actually costs $200 in SAFE resources utilized.


#40

The assertion is that larger payments can afford to cost more transactions. That is, they are not required to be instant as would be desirable for small transactions.

Edit: To add, remember that these aren’t costly blockchain transactions. These are relatively cheap close group transactions, where only the owner is changed.