Update of Gen1 NP Remuneration

[katiep]

Hi David F! Whatever node providers choose to do with the extra servers is something that DFINITY can’t be involved in.

By the end of the 48 months, you will have received the promised multiplier of your purchase price for those servers and therefore a return on what you spent on them. It is true that some NPs who owned 70 nodes would have to exit more nodes, but we do not consider this to be a penalty. Those NPs also will have received a far greater profit running the nodes over the course of the 48 months. 2.5x of the full cost of 70 nodes is a much greater profit than 2.5x the cost of 28 nodes. Plus you own the servers still, so you could certainly sell them to anyone to recap more of those costs. As far as another node provider onboarding them, that would be their business decision based on whether the rewards for Gen-1 servers fit their business model. The only problem is that we do not anticipate the network meeting more nodes for quite a while. As the topology posts explain, the target number of nodes can handle the growth that we expect the IC to have over the next year and beyond.

Regardless, the sale of your servers is not something that DFINITY can be involved in. That would have to be a business deal between those two businesses directly.

[katiep]

Hi David B!
Type-0 nodes are no longer usable on the IC. If the storage was removed, the nodes would no longer be healthy. The entire IC was upgraded to require (and use) the extra storage.

[katiep]

Hello David M! I will answer a few of the questions that I feel able to answer clearly, but I will ask others from DFINITY to chime in as well for other questions.

@DavidM

What is the planned timing for the proposal to go live on Gen 1 node provider interim rewards?

As stated, these interim rewards will not kick in until the original 48 months are up for each Gen-1.a NP. For a few, this is 11/30/24 (so the final reward distibution for them would be the Dec 2024 distribution), and for the rest, it is 1/31/25. Thus, there is a lot of time for NPs to make plans.

@DavidM

What is Dfinity’s perspective on the acceptable total number of Gen 1 node machines operating in the 2025 timeframe?

This is fleshed out in these two forum posts:

IC topology series - Node diversification part I

TL;DR The Internet Computer (IC) operates on physical node machines, commonly referred to as “nodes”. To achieve robust decentralization, these nodes are operated by different node providers, in multiple data centers, by different data center providers, and in many countries. At present: Decentralization is well-achieved in “data center” and “data center provider” characteristics. While “node provider” decentralization is nearing optimal, some providers operate too many nodes relative to avail…

IC topology - Node diversification part II

TL;DR This post explores metrics and models for node decentralization and presents strategies for future node onboarding and offboarding decisions. Metrics: A recap of key metrics for node decentralization, in particular the Nakamoto coefficient. Modeling: Introduction of an optimization model designed to reach decentralization targets with a minimum node count. Community involvement: After incorporating community feedback we aim to proceed with a motion proposal for NNS agreed rules for node …

@DavidM

[DISCUSSION: We have 1100 current machines and are short type 2s. Bjoern’s analysis suggests a need for 750. How many Gen 1 node machines does Dfinity think need to exit the network? Are you OK with all current Gen 1 node machines staying online in the network? Or do you feel that some number needs to exit? As pointed out in DavidF’s note, several node holders have more than 28 and a few have more than 42 - his question about selling nodes might have a different answer if you are trying to get nodes to exit. ONE NOTE: Bjoern’s analysis did not seem to factor in downtime and nodes offline (significant and growing from what I have seen) - I think all of the numbers need to be increased for cushion/slack]

We believe that 750 is all that is needed, and this DOES include the need for spare nodes and nodes that are down.

It is clear that, for tokenomics needs, the community wants to see extra nodes exit so that rewards decrease, thus helping the token price to increase. The hard cap of 42 nodes per NP will result in a minimum of nodes exiting. The rewards scheme being proposed will reslut in “extra” nodes beyond the 750 which remain on anyway costing the IC less than the nodes that the IC actually needs.

@DavidM

What is the objective of the decreasing marginal return remuneration scheme - would not a flat payment scheme for at least the first 28 nodes be simpler and accomplish the same objective?

This is a reflection of the common desire for node rewards to reflect the needs of the IC as much as possible. The IC needs NPs to own fewer nodes. Thus, a remuneration calculation that rewards NPs more for a smaller number of nodes is more reflective of this than a flat calculation.

@DavidM

What does Dfinity want to have happen for node providers hosting more than 28? More than 42?

The servers are owned by the NP, not DFINITY. DFINITY does not have any authority or desire to influence what any NP does with their extra servers.

I will ask some of the others to reply to some of your other questions.

[DavidB]

Hi Katie,
I wasn’t aware of this. So all the node providers which didn’t buy the storage in the beginning bought it later on?

[katiep]

In one case, a NP split the cost with someone else so rewards are split, but yes, all the NPs bought the extra storage and it was installed in the entire network a year or more ago. The IC is using all the drives now, so if a hard drive fails (which has happened a few times), it causes the node to go down until that drive is fixed or replaced.

[SvenF]

@DavidM

[DISCUSSION: I think a flat rate up to 28 (again, maybe somewhat higher for slack) would be simpler and fairer. There are a couple of subnets where a node provider with 42 units would not harm the Nakamoto coefficient. However, the proposed remuneration scheme suggests a negative operating return above 28 - which would mean that any node provider with more than 28 would shut down those machines. If the network can accommodate a few node providers up to 42 without impact, why not just extend the curve to those nodes as well? Similarly, I had some questions for a node provider with over 42 nodes’ impact on the Nakamoto coefficient. Do extra machines hurt the Nakamoto coefficient or do they just not add any value to decentralization? If fairness is a consideration, and the extra machines do not hurt the network maybe an answer would be just to extend the payment curve for the 2 or 3 node providers that fall into this category. Or, perhaps these excess units could serve as a backup pool for the rest of the network]

Hi @DavidM regarding the impact of larger NPs on the Nakamoto coefficients: if these nodes would be in a subnet, yes then they would have a negative impact on the Nakamoto coefficients, most likely even if these nodes were kept as spare nodes that are going to be swapped into a subnet when another node fails. So yes, these nodes do not harm the network but they do not add any value to the IC network as well in terms of decentralization, unless in the future the target topology is increased and more node machines are needed.

[SvenF]

@katiep

We have created a wiki page outlining an interim remuneration proposal: rewards for Gen-1 node machines after 48 months.
https://wiki.internetcomputer.org/wiki/Proposed_Gen-1_Remuneration_Model

Hi all, just making you aware the motion proposal is live for voting see proposal 127044.

TL;DR This proposal is for introducing node rewards for Gen1 node machines beyond the initial 48 months. In line with the target topology and with node rewards for Gen2 node machines, the reward mechanism are optimized for 28 node machines per node provider and have a reduction coefficient for each additional node.