I hate seeing someone fall for the climate change meme so I would like to invite you to re-open your mind and play with the idea you have been had. A lot of countries already force their citizens to pay a climate/carbon tax so in the eyes of our rulers who mind you only allow us to use their “approved” power sources the “problem” is long $olved, don’t forget to tip your taxman either.
If the programming is buried too deep within you mind and you cannot let go of their incepted ideas then rest assured you will feel mighty happy when carbon credits get brought in.
Thank you for acknowledging this is a complex problem and for sharing concerns regarding gathering data.
Similar concerns are acknowledged by the authors of “The Energy Footprint of Blockchain Consensus Mechanisms Beyond Proof-of-Work” published on April 4th 2022. This study has good provenance, being a collaboration between University College London, Imperial College, Kings College and University of Bayreuth. The concerns, including data integrity and hardware configuration, are approached with appropriate scientific best practice, “We approach this research gap by formalizing a basic consumption model for PoS blockchains.” https://arxiv.org/pdf/2109.03667.pdf The abstract is quite short and is definitely worth a speedy read.
The above study compares Ethereum, Algorand, Cardano, Polkadot, Tezos and Hedera.
This Jan 2022 report by Crypto Carbon Rating Institute looks at Total electricity consumption [kWh/year], Electricity per node [kWh/year], Electricity per transaction [Wh/Tx] and Total carbon emissions [tCO2e/year] for Cardano, Polkadot, Solana, Tezos, Avalanche and Algorand. No single blockchain out performs all the others across all criteria, which isn’t surprising https://www.carbon-ratings.com/dl/pos-report-2022
The conversation around energy consumption across all aspects of IT is only going to grow. Therefore, finding ways for the Internet Computer to participate in these discussion is imperative, or once again it will be notable only by its absence.
@icpjesse Hold up there I don’t think we solved this whole part of what sustainability data DFINITY thinks it can feasibly get. I have a few questions/potential solutions for @Luis
I understand that it’s hard to understand the full impact, but hopefully we can start with a few basic metrics that are based off two things.
Power consumption from the Boundary Nodes (which I understand are 100% DFINITY controlled atm)
Emissions reports from the data centers that are running nodes on the IC.
This way, DFINITY can present boundary node power consumption estimates with near ~100% certainty, and can provide 3rd party metrics as well. Those 3rd party metrics may be inaccurate, but the goal would be to start by trying to provide at least one or two metrics that are accurate to within an order of magnitude.
I would imagine DFINITY would want to use this information to market the IC, and the more accurate the data from data centers becomes the better. It might even make sense to incentivize the data centers that provide accurate power consumption with additional rewards for the time being.
Just FYI I am moving ahead with getting the proposal submitted to the NNS. My reasoning for doing so is two fold:
All of our action items are “Ask DFINITY…” so they are nonbinding and the answers to these questions will need to be solved by DFINITY no matter what. I’m optimistic it passing will help management decide to allocate further resources to help answer the questions more fully.
It’s very likely that the cost of having an NNS proposal fail will dramatically rise in the near future. The fate of this good-faith proposal is unknown, but given the controversial nature of it, failing is of course a real possibility. Given how much discussion here has slowed, I’m motivated to get it submitted. I expect that once live on the NNS this thread will become significantly more active.
By design, this proposal leaves everything in the hands of DFINITY for execution - we are just trying to set a direction and voice this need for another marketing tool (plus save the planet, etc).
Sorry @icpjesse for the delayed response. We had a long easter holiday in Europe.
I tried to explain that DFINITY can’t provide such details because we don’t have them. Technically the node providers won’t be able to provide that details without installing additional hardware. Even if they would do that these very precise metrics wouldn’t allow us to draw any conclusions about the sustainability of the IC.
You could do an estimation of the current total power consumption of the IC with details that are already publicly visible: The number of nodes per data center and node provider you can get from the ICA’s public dashboard. This gives you a number of active nodes per DC. There you can also find the DC provider/owner and look for their carbon footprint. On the ICA website you can find the specs for the current dual-socket AMD Rome nodes. With these details you can calculate a total power consumption of the IC under full load and how much of the power is provided by DCs that claim to use renewable power sources. It still doesn’t say anything about the sustainability of IC.
Drawing conclusions to sustainability based on such energy source and consumption details, nevertheless how precise they are, is what I would call greenwashing.
Wasting power stays a wastage nevertheless where the power is coming from and wastage is never sustainable. What the power is used for is therefore relevant for the sustainability in the very first place?
Let’s assume we need to run something that is sustainable by nature. Let’s say a car pooling or a second hand exchange portal. The sustainability could then be reduced to the protocol efficiency and its energy source. A comparison between a legacy LAMP setup and the IC would probably already show a higher efficiency if you take the redundancy and geological distribution of a subnet into account. But that’s hard to prove and we are improving the protocol efficiency constantly.
But now let’s assume we need to run something that can only work on-chain. Let’s say a voting machine providing the first democratic elections in an autocracy or an NFT portal giving the exploitation of the rights back to the artists. Here you would need to compare the IC with other blockchain projects where the technical edge of IC would most likely show a higher protocol efficiency.
Referring to 1.
Boundary and replica nodes have almost the same power consumption. The utilisation and efficiency is different though. A boundary node is basically a nginx reverse proxy that translates http to ic and back. 10 boundary nodes are currently serving about 600 nodes. They could serve much more nodes and subnets but the redundancy and geographical distribution is asking for more boundary nodes not the performance. That’s why soon every node provide could run one or more boundary node based on the traffic in the region the DC is located.
Referring to 2.
As explained before DFINITY doesn’t have this information. The node providers do and everybody who wants to know that can find this out by themselves based on the information that the ICA is providing. And again these reports would probably only be trustworthy for some us. They can’t be fully verified.
I don’t see any open questions nor something that exclusively DFINITY would be able to answer.
Everything you were asking for can be evaluated by anyone. The necessary information is provided by the ICA (see post above).
The IC is a decentralized system where exclusive powers and knowledge should not exist.
Europe had four free days. That could explain the slow down. And perhaps the fact that there isn’t much to discuss.
I would say that this goal is already achieved and for my understanding your proposal only seems to focus on the environmental aspects of sustainability.
The IC isn’t DFINITY. DFINITY makes a big part of what we currently see as the IC, but not all of it. There are many other entities involved.
As I explained above someone outside DFINITY could start modeling scenarios and thereby prove the sustainability of the IC in all aspects for those who didn’t understood that yet. For the sake of credibility this shouldn’t be DFINITY but some independent entity.
What if we just started with overall power consumption and not overall efficiency? Efficiency sounds a bit harder.
I’m assuming there weren’t always 60 nodes per boundary node, so it might be interesting to look at how power consumption has gone up over time. I also assume we can also look at metrics like overall traffic coming into the boundary node and compare that with the power utilization on a daily, weekly, or monthly basis.
Totally agree, would it be fair to say that a node provider could benefit in some way (grant/community fund) from investing the time to get this information?
@icpjesse Maybe the next step here then is to get in contact with a node provider and try to make this happen?
Because the specs are on the ICA website. Load and specs are almost the same.
And leave sustainability also out?
I previously explained how the total power consumption under full load can be calculated. You could propose an average load over all subnets metric on the public dashboard and by that add the power consumption. But it would be interesting to know why (if you want my vote). What do you want to derive from the power consumption?
A positive branding and positioning for the IC in the context of sustainability just based on power consumptions, estimations and trusts could be seen as greenwashing.
That’s why this needs the social and economic (efficiency) aspects that I previously explained and you initially proposed.
This just needs much more research and refinement. I’m personally a huge fan of such ideas. Not necessarily by teaching others but just by proving that there is a sustainability by design in the IC.
So then we can simply do something like multiply the power consumption of a single IC node multiplied by the (number of replica nodes + number of boundary nodes) and now we have a very rough estimate (to within an order of magnitude) of the power consumption of the IC infrastructure.
Then we can make a similar comparison against Bitcoin/Ethereum, and can maybe postulate on what the overall power consumption of a fully scaled up IC would look like in the future. Would it be equal to that of a large city, a small country, etc.?
I wouldn’t want to make any blanket statements if you don’t feel confident extrapolating some of the power consumption stats like this, but if we can provide an error boundary of within an order of magnitude - that has to be at least as accurate as many of the 10-20+ year climate forecasts we get from the IPCC, right?
I think this is where we disagree. I would define “Greenwashing” is providing a false narrative about sustainability practices - empty marketing with no real action. What IC (and you!) have done is design a very efficient system, one worth talking about. It’s not greenwashing if we really are taking a sustainable action. I want to know the details of what we are already doing, and promote that.
If your position is that any marketing related to sustainability is “greenwashing,” then I think we disagree on the premise, and would agree you should not vote for it. But I see failing to do so is a self-goal when it comes to marketing the IC.
“A node consumes about 500W in idle state and about 700 W under full load. App subnet has 13 nodes, and can handle 750Tps, so 0.000000123340977 kWh per transaction under full load.”
This comment was posted by a Dfinity researcher on social media following an IC energy consumption claim by a third party, which clearly deserved a bit of challenge/clarification.
This proposal should ensure third parties are able to make informed comments without relying upon the happenstance of a knowledgeable Dfinity Foundation member stumbling across and responding appropriately to poorly informed reporting.
There we agree. I think the disagreement we have is on the sustainability practices. Measuring the consumed power and comparing it with other crypto projects without any relation to use cases, workloads or social benefit I hesitate to see as a practice that proves the sustainability of the IC.
The power consumption and total environmental footprint is needed though for the real actions.
Happenstance? I know about that tweet. BTW Jan isn’t just some researcher but the CTO of DFINITY.
I provided the power consumptions details. He asked me to respond to this thread that Jesse linked in his response. Not sure if you read the comments of this tweet. This isn’t about sustainability but some abstract comparison of coincodex that used obviously wrong assumptions about the energy consumption of something that you could call a transaction on the IC. Dom’s reply shows how uncomparable the IC actually is. The coincodex article that Jan is revering to shows that even better. Pulled out of that context Jan’s tweet could be understood as greenwashing, but what coincodex tried I think is very near to what Jesse is trying to achieve. What I’m trying to say is that the precision on the energy consumption of IC, even with the correction that Jan tweeted, isn’t sufficient to draw conclusions about the sustainability of the IC.
@Luis I’ve been following Dfinity since 2016 and have participated at many levels over the years. ::edit removed content::
We are all on the same side and shouldn’t be sniping at each others’ assumed knowledge and understanding, but focus on the proposal, which is to provide some level of transparency around energy consumption.
In this reddit post my colleague Alin tries to calculate the power consumption per transaction. It shows how many variables need to be estimated and how poor the comparison with other projects still is.
That post is what inspired my original Medium article about ICP energy usage, and submitting this NNS proposal.
Thanks again for your engagement on this issue Luis - we might not agree on everything but I really appreciate your thoughtfulness on this topic and work on ICP.
