Thoughts on current ANT upload pricing and long-term sustainability

Development
1

I wanted to open a discussion about the current upload pricing on Autonomi and how others in the community view it.

Over the last two weeks I uploaded close to 1 TB of data. The effective cost for that was roughly:

  • about $200 in ETH fees

  • about $0.10 in ANT fees

That means 1 TB of perpetual data, effectively a permanent burden on the network, was stored for ten cents in storage cost. I understand that right now capacity is high and demand is still growing, so pricing is expected to be very low and to find equilibrium over time. That part makes sense.

What troubles me more is the lower bound of pricing and the lack of clarity around what ultimately determines the ANT cost in practice. The mechanism is described at a high level, but it is still difficult, at least for me, to reason about where a realistic long-term price floor might emerge.

I also wonder whether extremely low storage costs could create incentives that are not ideal in the long run, for example very large uploads that occupy resources for decades while contributing very little to the long-term sustainability of node operators.

One idea I have been thinking about is whether some form of soft lower bound could make sense, not necessarily a fixed price floor, but perhaps a guideline where storage cost per GB does not fall far below real-world storage economics. On the other hand, I can also see arguments for letting the market fully determine pricing, especially during the bootstrap phase of the network.

I would be very interested to hear how others think about this:

  • Do you see the current pricing as purely a temporary bootstrap effect?

  • At what level of network utilization would you expect pricing to start reflecting real long-term storage costs?

  • Do you think extremely low upload costs pose any risk to the network, or are they simply a feature of an early-stage system with excess capacity?

Curious to hear different perspectives.

9 Likes
2

I brought this up yesterday too, I am of the opinion it should be adjusted.

It can remain low without being essentially zero.

If a minimum network size is required, it should to be taken into account, not only storage availability.

Idk how feasible it is, but pay more to support nodes while the network is under a certain size is my thinking.

2 Likes
3

Quick update. I was looking through the whitepaper to find the pricing graph I was looking for, but I actually found a different graph.

image
image841Ă—470 32.6 KB

Does anyone know if this pricing graph changed recently (last year or so?)

As far as I can remember, the graph had a midpoint at 50% pricing wise and to both extremes (0 and 100%) the pricing responded parabolically.

I still think that pricing should never be as cheap as it is now, but I do think this graph makes more sense with the price increases we’re currently seeing (extremely slow).

What do you guys think the equilibrium sweet spot is for storage used? I’d say it would make economical most sense if you settle somewhere around 70% storage used. This is enough wiggle room to support large storage inflows while not wasting too much unused resources.

1 Like
4

Wow, that’s very concerning. Seeing the disparity between ETH and ANT highlights a fundamental issue that I hadn’t considered.

Any adjustments in ANT pricing are swamped, insignificant and will have little effect on the demand side.

I don’t plan to try and figure out what this means in practice but it is so far away from the conditions used to determine the current design it will likely need a rethink.

Even if you are paying everything in ANT, the problem remains.

5 Likes
5

Yeah I agree. The graph I ran into now does make a bit more sense. I thought we’d see the largest price hikes in the very early stages (0-1% is a larger increase than 1 to 2%). However, the graph I posted just now indicates we wont see major increases in ANT cost until the network approaches over 50% storage capacity used.

4 Likes
6

For what it’s worth. I’ve reduced my node count from 200 to 50 nodes just now and have doubled the upload speed. I’m laser focused to fill up the network to at least 15GB/node.

6 Likes
7

This was largely without Merkle, was it?

8

I’ve killed all mine and my VPS to save money so I can still afford whisky. But I’m not sure £600/yr is going to be enough :man_shrugging:

2 Likes
9

I think around 70% is with merkle, 30% without. Edit: Large amount of the merkle was before they reduced the data size for merkle payments though, not sure if that affected much though.

2 Likes
10

That’s crazy cheap for 1TB but also crazy expensive on fees.

It really baffles me how the network will maintain itself with all this data that doesn’t need to be paid for again by the owner of that data. It seems like it would be better to at least have to pay a small amount to retrieve your data each time, maybe even to maintain it? I really hope the pay once approach works out because it is obviously a huge selling point, but it doesn’t make sense from a free market perspective to me.

4 Likes
11

What’s great now though is that we have a real economy and just see how this evolves. I’m keen on seeing network shrinking and nodes filling up.

The ETH fees are obviously a bad thing, but it’s not very clear how bad when talking about big Merkle uploads.

2 Likes
12

Pardon the bad graph, but didn’t want to spend too much time on it so had AI draw something up. It does represent how I think upload price should work though:

image
image790Ă—470 24.2 KB

A rapid increase in price from 0 tot 20% capacity used. followed by the current price increase graph to make sure the network will eventually settle around 70% usage (again, not wasting too much free capacity while maintaining wiggle room for larger data inflow). I think from a economic point of view it would make a lot more sense to allow cheap uploads when the network is extremely low in usage (extreme cases), but rapidly increasing to make sure not too much garbage is uploaded for free (and/or price manipulation by increasing node count to lower usage and then upload).

In the grand scheme of things I do think that we’re at such an early stage and allowing 70 TB of data at a extremely cheap price isn’t the worse. Eventually the network will settle at a economic point that makes sense for node runners. I’m personally hoping we get there sooner rather than later (which is why I’ve reduced my running node count by 75%)

2 Likes
13

Right now, 256 chunks cost around $0.12. If I remember correct that’s supposed to be for 1 GB, I think. That’s $12 for 100 GB permanently (in just ETH fees). I don’t think that’s the worse and I also believe it can be optimized further. Right now, I do believe there are some inefficiencies when it comes to different files, I recently uploaded a bunch of smaller files (within a folder), and it made payments for each file separately (4 chunks each), it did that a thousand if not more times. That’s extremely inefficient, but I assume can be optimized as well.

2 Likes
14

This doesn’t seem good, though it is of course temporary. I don’t like the huge imbalance between ETH fees and ANT fees, but don’t imagine there’s a simple mechanism that could ensure the ANT cost of upload is e.g. equal to the ETH cost in total value.

Once ETH fees became part of the equation, I would probably have kept the store cost curve the same, but with an added ANT floor price. Even if the floor price was low, but made the ANT fee make up at least, e.g. 10%, of the total fee to upload, it would have been better and more rewarding to node operators seeing useful earnings coming in, and better for uploaders knowing they’re actually paying network nodes, not just ETH stakers.

I think so, and it seems quite positive to me.

The value of uploading to Autonomi is lowest while utility of the network is lowest to users, which is at the start, so it makes sense for the price to be lower. These early data uploads are also valuable for testing the network in practice with real data.

As utility improves, I expect a steady level of demand for uploading will form. Then, nodes will fill more quickly until payments are sufficient to warrant new supply (nodes) joining, forming some kind of balance with node operators responding the the level of demand.

I don’t see any risk to the network assuming the network does see a rise in demand in the not too distant future.

If demand does emerge and the network rapidly grows in the coming years, the early cheaply uploaded data will become such a tiny portion of total data stored that it will have pose zero risk to the network.

The long-term risk to the network I see is not from the pre-demand bargain uploads, but the eventual increase in upload cost that will occur when the network growth rate slows as it approaches market saturation. At this time, the ratio of old to new data held by nodes will increase, pushing up the data overhead of running a node, hence increasing the cost of running a node.

Hopefully by that time, the scale of the network and utility offered by network effects and the richness of the ecosystem using it will offset the steady cost increase to find a good equilibrium that doesn’t cause a doom-spiral of cost increase → lower demand → lower node revenue → node exits → cost increase → lower demand etc etc.

That’s some of my thoughts for what it’s worth. I’m interested to see how the team’s thinking on economics / tokenomics has progressed with Autonomi 2.0.

5 Likes
15

Also the dynamics will change a LOT once there is a lot more data on the network and merkle batch size increases again. (fees down)

The reason is the pricing curve relies on the nodes having reasonable amounts of storage and not the “Oh this is cute a few GBs of my disk is being used”

When its more like "This is actually serious stuff here" then the curve will do as is desired

To artificially raise the ANT price to satisfy a perceived lack of cost when the network is at the cute project stage is to invite similar criticism in the opposite direction once the nodes are at 20% or 30%

The idea was always to encourage people to upload so that the network is out of the cute stage and people have to actually have to be providing something real.

And do I really have to remind ourselves that this network is not as a money maker, but to compensate people for providing up their spare capacity. If seen as a money maker then the curve would indeed be different and the criticism would be why so expensive in ANT costs

1 Like
16

Thank you @mightyfool for doing the uploading and putting this together.

Those figures are shocking to see on 2 fronts:-

  • The vast different between the spend on ETH and ANT. It’s a real shame we have to use ETH.
  • The incedible cheapness of uploading 1TB of data for storage forever (or the lifetime of the network anyway) and free retrieval even with the ETH cost. If there is a cheaper way of doing that I’d like to know about it. This is still the deal of the century. The ANT price deserves to be 1,000 times higher with or without ETH being needed.

(We’ve talked about the dependency on Ethereum in the past and that as well as the extra cost the network is beholden to another system continuing to work.)

As regards tweaking the thresholds to take into account the Ethereum cost I thought we were all supposed to be fans of the free market? I can see value in changing the curve though similar to what @mightyfool recommended.

The most important point though is @happybeing’s about the disparity between the ETH and ANT costs. I didn’t realise it was that bad!

A 1% change in the price of ETH makes more difference to the total cost of uploading data than a tenfold change in either the price of ANT or the cost to upload a chunk or both at the same time. That doesn’t just skew the economics of the system as a whole, it completely capsizes it. Unless someone much cleverer than us knows something we don’t that is a giant problem.

Edit
Maybe that’s a bit harsh. It only really directly affects the upload side of the equation. If people have a choice about when to upload they will watch the ETH price rather than the ANT price. It doesn’t affect the economics of running nodes except in that the whole economics are affected by whether people upload or not and when.

2 Likes
17

Is this not a little contradictory.

  1. The network needs to charge as little in ANT as possible for as long as possible resulting in little incentive for “most” to run nodes

  2. To get the node numbers up, have emissions and let it rain.

The fact that we had emissions was saying that incentive is required to bring outsiders in but the obvious incentive to me is upload costs being nowhere near zero in ANT.

4 Likes
18

Yes.

The project for such a long time was to provide the services on a minimal cost basis where node runners were compensated fairly for their supply of resources. A moving away from the trend of charging the highest price people are capable of paying for storage.

The emissions came from a place where the excess token had to be distributed and the maximalist approach of just distributing it at a set rate while the node runner numbers were tiny was not the best way was it. Maybe distributing it according to %age data stored would have made a lot more sense since it is rewarding maintaining nodes/storage rather than just running software. And rewarding uploading good/garbage data without control over quality of data will be similar silliness

3 Likes
19

This is destroyed by the ETH fees.

3 Likes
20

Well they are still trying to reduce those.

But yes the disparity and lead weight it has made is a huge issue.

4 Likes