Optimizing Farming Rewards : Farming=Quality+Quantity

#1

Hi TFarmers,

If we start with the Game Theory hypothesis that farmers are rational agents and that their primary activity is to farm tokens by offering hardware resources on the Grid, we might conclude that farmers are seeking to maximize farming rewards.

As of now, with the current tokenomics, farmers will get rewards from proof-of-capacity: what they offer to the Grid. But when it comes to proof-of-cultivation, farmers will not necessarily get rewards when users use the Grid. To get rewards in this case, one would need to offer sales channel, etc. (This is a topic in itself, with great discussions on this Forum.)

We also know that farmers are often aligned with the Threefold ethos of building a new Internet for everyone, while also offering the best product on the market.

With the above, we have that farmers have, on the economic side, more to gain by offering a lot of resources (quantity), without having to look at the “quality” of the resources, as this does not translate as directly to TFT farming rewards.

Examples :

  • 1TB SSD SATA farms as much as 1TB SSD nvme. NVME is more expensive and of higher quality.
  • DDR3 vs DDR4 ram: 32GB DDR3 or DDR4 farm the same amount of TFT. DDR4 is faster and consumes less energy. DDR3 is cheaper.

So my point is:

With the current farming reward that is only precise up to the quantity of resources (X GB of RAM, Y Threads, etc.), but not yet up to the quality of resources (DDR4 > DDR3, SSD nvme > SSD sata), rational agents aiming for economic gains will aim at quantity > quality.

As a result, it would translate into sub-optimal Grid resources, quality-wise.

Here’s a proposition:

By refining the TFT farming rewards, (how) could we find the “perfect” combination of farming parameters that would incite farmers to invest in HIGH QUALITY resources, and not just aiming at having a high quantity of resources?

The obvious answer is to include quality in the farming rewards, thus to distinguish the quality of the hardware/resources that farmers offer on the Grid.

What do you guys think?

Do you have any suggestions?

What would be the best parameters to check?

Happy Farming!

1 Like
#2

Hi @Mik. There is a portion of “proof of capacity” which is IPv4 and bandwidth driven, that only farms reward when these resources are used. Remember, to be a utility token the token can only come into existence by creating a utility

1 Like
#3

This is true, but if I may, I also want to present a different angle to this topic. Speed of components / 3nodes is important just like bandwidth, cpu frequencies etc. Mean Time Between Failure of component is a different quality aspect, which is covered by the farming model today, as a failure of a component leads to a part of your farm no longer being functional and fewer tokens will be rewarded if not repaired. So the basics are covered.

The other statement is that certified capacity needs to have a couple of stringent requirements to be certified, one being a locked-down BIOS and a signed zero-OS version to boot. In this category, we will have certified hardware providers that will create these certified 3nodes and we should put requirements like this into those servers as they are eligible for a bonus in regards compaired to DIY.

So - we have some of the basics covered - and I think there is an incentive to take the best components one can afford for DIY which make it inclusive and equal for anyone to come become part of the grid in your own unique way?

3 Likes
Farming reward verification tool
#4

Thanks for the replies. I wrote some things but it was way too long (hence the delay in writing). Here’s a shorten version.

I completely agree that TF has the basics covered.

But just as we have an ongoing conversation about CPU and passmarks, I was wondering if this kind of thinking could, and if it should, extend to other components of a 3node.

So for example, why would a farmer install a NVME SSD if they can put a SATA SSD for cheaper. Same for DDR4 vs DDR3 ram.

And, most importantly, would a TF user notice the difference in the actual Grid experience? NVME vs SATA, for example, but higher quality component vs lower quality component, in general.

In the same line, would a user see the difference between using a 3node that is recent vs a 3node that is 6+ years old?

Said simply, if a user doesn’t notice the difference, than it doesn’t if farmers decide to go with less expensive hardwares with the same specs (e.g. DDR3 16GB of ram, instead of DDR4 16GB of ram).

On the other hands, if users will notice the difference, how can we incite farmers to go for DDR4, NVME, etc. instead of lower quality version of the technology (DDR3, SATA, etc.)?

I hope that makes sense!

Many farmers ask if they should go DDR3 or DDR4, obviously the easy ROI-based answer is DDR3. But what is best for the Grid? Will users notice the difference? That’s my main question.

Thanks!

#5

I assume that somewhere in the future there will be a “performance” tier.

I don’t think that you will notice the differences for a simple K8 cluster or owncloud hosting. I think you could solve part of this with fairly simple metrics in

  • CPU performance (although we would have to measure performance per core when the whole CPU is under full load. So passmark divided by cores)
  • memory bandwidth
  • SSD bandwith (buying crappy SSD’s is something farmers will deal with themselves in terms of degradation)
1 Like
#6

These are interesting questions. I think it’s fair to assume network bandwidth will be the bottleneck to performance far before and sata/nvme protocol limits are hit? If the workload is, say, data backup and we assume host and client both have a gigabit connection - this translates to a theoretical bandwidth limit of 125 megabytes/sec (bytes = bits/8) so in terms of whether a user will notice performance difference between sata (~500MB/s) and nvme (~1500MB/s+), the answer is no; upload/download speed will obviously be limited by connection in this instance. Likewise with compute - what good is having a bleeding edge processor and top tier memory bandwidth if you can’t take full advantage because of a network bandwidth limitation. We can expect some level of hardware performance disparity as network speeds increase but given how fiber is still not an option in many places I don’t expect that anytime soon. Just my 2c.

1 Like