What is proof-of-service?

[amanda_b_johnson]

In researching this question, I'm quite surprised to say that Google has only shown me two results that have "proof of service" in their titles: this and this.

So, really, what is it? And how does it differ from PoS and/or PoW? And, most importantly, why does it matter to Dash?

Thanks.

 

[qwizzie]

The way i understand it Proof of Service / PoSe currently runs on deterministic masternodes and gives masternodes in Core quorums that are misbehaving PoSe scores and ultimately PoSe bans them.

With regards to Dash Platform (which is basically a sidechain) the Proof of Service in theory makes sure that nodes inside a Platform quorum can be monitored and punished, if they misbehave.

Dash Platform's sidechain runs on a fork of Tendermint, which is named TenderDash. Which is a PoS (Proof of Stake) blockchain with a deterministic protocol.
A robust PoSe scoring solution for a Proof of Stake blockchain has not been developed yet. By anyone.

A flaw of Proof of Stake blockchains for Dash is that if a large masternode entity stops its Platform nodes all at once, either due to an upgrade or due to an attack, there is a chance that the nodes of that large entity can stop a Platform quorum, which will then cause the Platform chain to stop. If the Platform is supported on all masternodes (which is the orginal vision of Dash Evolution), then that will affect all masternodes on the Dash network.

This triggered the discussion :

Source : https://www.dash.org/forum/threads/...gh-performance-nodes.53374/page-3#post-232232

Proof of Service / PoSe scoring solution for Dash Platform in theory seems to provide a way to mitigate this and it is my understanding that this was also to be developed for Dash Platform initial release (i recall having it seen on the Dash roadmap or seen this feature mentioned by devs in the past).

The lack of a robust PoSe scoring solution for Dash Platform brings some risks to the launch of Dash Platform, which the devs seem focused to mitigate by introducing higher collateral+higher hardware requirements, while sacrificing decentralisation. Dash masternodes will end up with two groups, one larger normal masternode group providing Core services and one much smaller High Performance Masternode group (4K or 10K) providing Core services + Platform services.
If the much smaller High Performance Masternode group go down with their Platform services + Core services, the larger normal masternode group with its Core services is not affected.

With regards to the other solution (Platform on all masternodes) with increased hardware requirements and no change in collateral : it is now the question if that can be safely used to startup Dash Platform, or if it causes problems during startup or during large updates. That solution would contain a 50/50 split of its masternode portion of the blockreward (50% as MN blockreward / 50% as Platform credits) and it would provide Core services and Platform services.

A robust Proof of Service scoring solution is pretty much a requirement, if Dash will ever want to get into sharding. Which should help with Platform Storage and the availability of that stored data in the long term.

There are also benefits to using Platform HPM 4K / 10K : very low Platform Storage fees (a few cents) and high transactions per second (TPS). It is the question if those are really needed for Dash Platform's initial release to Dash Mainnet. Optimization and fine-tuning can also bring lower Platform Storage fees to Dash Platform after launch.

The Platform on all nodes solution has higher Platform Storage fees and lower Transactions Per Second (TPS), but is decentralized and does not lock masternode owners out of Platform rewards, if they don't have the 4K or 10K collateral.

So my feeling is that Masternode owners are getting pressured by DCG to go for the 4K / 10K HPM solution, marketing mostly the low Platform Storage fees and high TPS, because as it turns out DCG did not develop a PoSe scoring solution for Dash Platform over the years, struggled with this Proof of Stake flaw for some time, now feels pressured to release end of this year and rather want masternode owners to make the decision on how to startup Dash Platform for them. Giving the Dash masternode owners currently some crappy options. Two of the solutions lead to centralization, one solution possibly has a safety issue.

The above is my own understanding which i could have misunderstood, an explanation from a developer would be appreciated.
It could very well be possible that the safety issue with regards to the startup of Dash Platform is less of a problem, then i currently think it is.

 

[xkcd]

There are also benefits to using Platform HPM 4K / 10K : very low Platform Storage fees (a few cents) and high transactions per second (TPS). It is the question if those are really needed for Dash Platform's initial release to Dash Mainnet. Optimization and fine-tuning can also bring lower Platform Storage fees to Dash Platform after launch.

This is the main reason the High Performance Masternodes was introduced to the network, the cost of replicating the data across 4000 nodes is way too high to make Platform economical to use, only running it on fewer nodes makes sense.

There are no further optimizations that can be done to reduce the data storage requirements by an order of magnitude, keep dreaming, but no need to give the community false hope in this regard.

The Platform on all nodes solution has higher Platform Storage fees and lower Transactions Per Second (TPS), but is decentralized and does not lock masternode owners out of Platform rewards, if they don't have the 4K or 10K collateral.

Purge your mind of this second class MNO notion, it has been explained to you for the umpteenth time and now you are plain ignorant. There will be a market between the MN types with MNOs freely available to move between the type pools, this will ensure the ROI in each pool is essentially the same. No MNO will be disadvantaged in the 4k or 10k plan, in fact quite the contrary, every MNO will get a higher ROI in either of these options, whereas in the 1k option all MNOs would be taking a haircut to their monthly profits of about $50 a month initially growing over time as more data is stored on Platform.

 

[vazaki3]

This is the main reason the High Performance Masternodes was introduced to the network, the cost of replicating the data across 4000 nodes is way too high to make Platform economical to use, only running it on fewer nodes makes sense.

Please, do not hide your assumptions, and talk with numbers.

IF the data stored in the Dashplatform are huge (huge=???), THEN the cost(cost=???) of replicating the data across 4000 nodes is way too high(high=???) to make Platform economical to use (economical_to_use = ???)

Of course you are not the Dash CTO, so you are forgiven for not presenting your belief in a formal way.....
But for the rest of us to be able to participate in the discussion, you have better use the formal way.

 

[denk]

in the 4k or 10k plan, in fact quite the contrary, every MNO will get a higher ROI in either of these options

This is not immediately obvious to me that ROI will be higher with the new HPMN mechanism. Can you explain in more detail, please - I probably still have my head squared in the current economic incentives, which I did not expect to change. I played with a spreadsheet I found on Discord (not sure it is correct actually) that calculates the different variables to the economic incentives of HPMNs. What is the key concept to the ROI growth that you claim? I guess if that is cleared out I have no big problem with the Platform upgrade based on HPMNs.

 

[denk]

the Proof of Service in theory makes sure that nodes inside a Platform quorum can be monitored and punished, if they misbehave.

Monitored for what exactly? Nodes in chainlock and instandsend quorums are monitored for the sake of participating in the quorum. So what is exactly the thing that needs to be monitored.

 

[qwizzie]

Monitored for what exactly? Nodes in chainlock and instandsend quorums are monitored for the sake of participating in the quorum. So what is exactly the thing that needs to be monitored.

The flaw of PoS blockchains for Dash is that if enough nodes in a Platform quorum stop all at once, it could stop a quorum, which could stop the Platform chain.
PoSe for Platform should then monitor participation in the Platform quorum, PoSe score and PoSe ban nodes if they misbehave in a Platform quorum and additionally auto reset the Platform quorum if it dectects stopped nodes there or if it mostly consists of banned nodes. This would keep the Platform chain alive.

I am no developer, and this is just my very simplistic idea of how PoSe scoring on Platform could possibly look like. There need to be some form of automatic reset for Platform quorums, to keep the Platform chain alive.

 

[xkcd]

Please, do not hide your assumptions, and talk with numbers.

IF the data stored in the Dashplatform are huge (huge=???), THEN the cost(cost=???) of replicating the data across 4000 nodes is way too high(high=???) to make Platform economical to use (economical_to_use = ???)

Of course you are not the Dash CTO, so you are forgiven for not presenting your belief in a formal way.....
But for the rest of us to be able to participate in the discussion, you have better use the formal way.

I didn't provide the numbers because they have already been quoted by QE elsewhere, but suffice to say the storage costs of data on Evo are directly proportional to the number of nodes on the network.

This is not immediately obvious to me that ROI will be higher with the new HPMN mechanism. Can you explain in more detail, please - I probably still have my head squared in the current economic incentives, which I did not expect to change. I played with a spreadsheet I found on Discord (not sure it is correct actually) that calculates the different variables to the economic incentives of HPMNs. What is the key concept to the ROI growth that you claim? I guess if that is cleared out I have no big problem with the Platform upgrade based on HPMNs.

If we look at the entire masternode network as a system where the baseline is what we have now, then tally all our costs of hosting we can give it some number, eg $20*3700. If we consider the option where everyone is forced to run Evo, the 1k model, then what we pay on hosting will increase dramatically, say to $50-$100 each driving down the ROI. However, in the 10k option, since half the rewards are distributed to those nodes, half the network will move there, but the hosting costs will be less, instead of 10*$100 per node, it will be $100 per 10k node, a tenth of the cost, since the overall cost of hosting of the entire network, ROI will be driven up for all holders.

 

[denk]

If we look at the entire masternode network as a system where the baseline is what we have now, then tally all our costs of hosting we can give it some number, eg $20*3700. If we consider the option where everyone is forced to run Evo, the 1k model, then what we pay on hosting will increase dramatically, say to $50-$100 each driving down the ROI. However, in the 10k option, since half the rewards are distributed to those nodes, half the network will move there, but the hosting costs will be less, instead of 10*$100 per node, it will be $100 per 10k node, a tenth of the cost, since the overall cost of hosting of the entire network, ROI will be driven up for all holders.

OK, I see now what you mean. You are comparing ROI when everyone must run Evo to the HPMN cases. That yes. But I was talking relatively to the current ROI we have today. The cost $20*3700 is perhaps what will need to be an average cost to run MN with the HPMN solution due to the requirement of running a fully indexed blockchain, but today you can easily run MN for half of that price or even less (as noted in the HPMN Economic Incentives Spreadsheet.)
To make it clear, compared to today, ROI will be lower with any suggested Platform upgrade, not higher! Argument that some case has higher ROI must clearly say higher than what exactly.
It also worth addressing the proposed 1k solution with everybody must run MN as both economic and technical nonsense, which DCG cannot seriously propose to the network or risk MN network's integrity.

There will be a market between the MN types with MNOs freely available to move between the type pools

Important correction: "... with SOME of the MNOs freely available to move between the type pools." Namely those who will have enough funds and can afford to use x-times 1k DASH collateral. We can expect the number of owners who can afford 4k and especially 10k DASH collateral will be order of magnitude higher than the number of owners who cannot so "freely move between the type pools".
Also recognize, MN owners, that the market between the MN types is quite complex to keep in balance. It does not depend only on number of nodes in each pool, but also on price of DASH, relative cost of running each type of node, or fees payed for Evo usage. Any change in any of these aspects imbalances the market. It is a perfectly legitimate to ask questions like: Do we want that? Do we need that? Or, cannot we avoid that?

 

[xkcd]

OK, I see now what you mean. You are comparing ROI when everyone must run Evo to the HPMN cases. That yes. But I was talking relatively to the current ROI we have today. The cost $20*3700 is perhaps what will need to be an average cost to run MN with the HPMN solution due to the requirement of running a fully indexed blockchain, but today you can easily run MN for half of that price or even less (as noted in the HPMN Economic Incentives Spreadsheet.)

Firstly, the low fees charged by AllNodes is due to the fact they only store one or few copies of the blockchain and run the nodes on individual clusters that host many nodes on the same hardware, ie centralised. The price can therefore be driven down and just $4.5/month hosting fees is completely unrealistic. Anyone that is arguing for decentralisation would not also be using AllNodes, because that is not decentralisation. I am personally paying $25/month per instance for a decentralised node.

To make it clear, compared to today, ROI will be lower with any suggested Platform upgrade, not higher! Argument that some case has higher ROI must clearly say higher than what exactly.

Disagree, because with the 4k and 10k models, while the cost of running the instance is higher, we run 4x or 10x fewer of them and that is where the savings come from.

Important correction: "... with SOME of the MNOs freely available to move between the type pools." Namely those who will have enough funds and can afford to use x-times 1k DASH collateral. We can expect the number of owners who can afford 4k and especially 10k DASH collateral will be order of magnitude higher than the number of owners who cannot so "freely move between the type pools".
Also recognize, MN owners, that the market between the MN types is quite complex to keep in balance. It does not depend only on number of nodes in each pool, but also on price of DASH, relative cost of running each type of node, or fees payed for Evo usage. Any change in any of these aspects imbalances the market. It is a perfectly legitimate to ask questions like: Do we want that? Do we need that? Or, cannot we avoid that?

Dash/USD is already a complex market place that takes in many factors such as BTC price, macro picture, DCG performance etc etc into account to determine the true price of a Dash. We all accept it as fair and we should also accept the new market place that emerges between the MN types as fair also. The data on MNOwatch.org clearly shows that there are enough MNOs with 4K+ collaterals to move in and out of the pools for 4k model and even 10k model to ensure the ROI between the regular pool and the HPMN pool will be close to equal at all times.

 

[denk]

Firstly, the low fees charged by AllNodes is due to the fact they only store one or few copies of the blockchain and run the nodes on individual clusters that host many nodes on the same hardware, ie centralised. The price can therefore be driven down and just $4.5/month hosting fees is completely unrealistic. Anyone that is arguing for decentralisation would not also be using AllNodes, because that is not decentralisation. I am personally paying $25/month per instance for a decentralised node.

Ok, I give you that today the average cost of running a MN is around $15 a moth, but not more than that. I would estimate the reality is even less than that, however I would agree with your arguments about decentralization in theory. The fact that you run one at $25 / month does not make it the case for others.

 

[denk]

Dash/USD is already a complex market place ... We all accept it as fair and we should also accept the new market place that emerges between the MN types as fair also.

Now that is a weak argument. It suggests accepting a complex market mechanism that will burden only MN owners (Dash internally) just because it can be fair (if MN types rewards are successfully balanced out). That is not enough of a reason. It has to have better ROI or some other sufficiently attractive value that can be transformed into improved ROI relatively to today. Otherwise MN owners cannot accept the upgrade simply because it does not make sense for them.

 

[denk]

... with the 4k and 10k models, while the cost of running the instance is higher, we run 4x or 10x fewer of them and that is where the savings come from.

Yes, but I have not seen the ROI being higher compare to today. That is a critical information to MN owners. So, let's show it, prove it, I would be glad if there is such case that shows with theoretical numbers and evaluations that ROI will/can improve over today.

 

[xkcd]

Now that is a weak argument. It suggests accepting a complex market mechanism that will burden only MN owners (Dash internally) just because it can be fair (if MN types rewards are successfully balanced out). That is not enough of a reason. It has to have better ROI or some other sufficiently attractive value that can be transformed into improved ROI relatively to today. Otherwise MN owners cannot accept the upgrade simply because it does not make sense for them.

I am not sure what you are trying to say here? Is it that you don't believe a secondary market will emerge between the two pools and MNOs will move between them in order to maximise their returns? Because if so, this would be the first time time someone that saw free money on the table and did not take it. Now where the balance point will lie, I cannot know, will it be 6.5% (regular) and 6.7% HPMN? Or the other way around? Or exactly the same? We can only guess, but my understanding is it won't be 6% and 7%, that is just too juicy for a MNO to not act.

Yes, but I have not seen the ROI being higher compare to today. That is a critical information to MN owners. So, let's show it, prove it, I would be glad if there is such case that shows with theoretical numbers and evaluations that ROI will/can improve over today.

Quantum has posted these numbers just recently.

At network start (no fees generated) AND Assuming the rewards even out with the market we get the following values..
All nodes run platform: 6.11%
1K split system: 6.49%
4K split system: 7.05%
10K split system 7.16%

At network start (no fees generated) AND Assuming platform nodes get slightly more rewards (educated guesses on how the system will stabilise).
All nodes run platform: 6.11%
1K split system:
Masternodes: 6.4%
HPMasternodes: 6.6%
4K split system:
Masternodes: 6.9%
HPMasternodes: 7.2%
10K split system:
Masternodes: 7%
HPMasternodes: 7.3%


At network start (no fees generated) AND Assuming platform nodes get slightly more rewards AND platform nodes run stronger hardware to better serve the network and ensure their profits(educated guesses on how the system will stabilize and hardware used).
All nodes run platform: 6.11%
1K split system:
Masternodes: 6.4%
HPMasternodes: 6.6%
4K split system:
Masternodes: 6.75%
HPMasternodes: 7%
10K split system:
Masternodes: 6.9%
HPMasternodes: 7.2%

 

[AgnewPickens]

Proof of Service is defined by the needs of the network,