The blockchain ecosystem boasts hundreds of different architectures, but there are two that are the most well-known: Proof of Work (PoW) and Proof of Stake (PoS). They also paint a simple, easy-to-understand landscape of energy consumption. PoW is energy-intensive, PoS is not. However, there is much more to this space.
At the DLT Science Foundation, we have been conducting research to understand the carbon footprint of Proof of Work and Proof of Stake. The core principle to calculate energy consumption is intuitive: identify the hash rate of the PoW network and multiply it by the energy consumption per hash, or if the network is not PoW, identify the number of nodes of the network and multiply them by the power usage of the average node. The core matter is investigating the characteristics of the hardware used in each case. To calculate carbon footprint, we also need to do research on the energy sources being used by the miners or the nodes. Nevertheless, there is a genre of blockchains that cannot be studied with either approach: blockchains based on storage.
Filecoin, Chia, and Sia are three blockchains with immense architectural differences and unique value propositions in each case, but sharing a common factor: they all have storage at their core. As a consequence, their energy consumption and carbon footprint does not depend on the amount of mining or hashing, nor on the number of nodes. Rather, it is a result of the amount of bytes being used. In this blog post, we investigate this a bit further.
The Common Threads: Sealing and Storing
Whenever a blockchain’s configuration is principally based on storage, there are two main factors that drive their energy consumption: The initial sealing of data and its continued storing thereafter. Sealing, also called plotting, is the process of encoding the data (and generating and submitting to the blockchain the corresponding proofs). This is the most energy-intensive step in the process, but it is a one-off expense. Once the data has been plotted, it must continue to be held and made available, and its integrity must be maintained. This is called storage and is a less energy-intensive activity, but also a continued one.
If calculating the energy consumption of PoW and PoS demands inquiring the energy consumption per hash, or per node, and to then identify the hash rate or number of nodes (respectively), for storage blockchains the investigation of the energy consumption per terabyte (and finding out the number of terabytes sealed and stored) is required. In this direction, the work of Alan Tensil and the Filecoin Green team constitutes a reference in the matter. At the DLT Science Foundation, we were pleased to see that the initial paper that outlined the methodology for the Filecoin Green dashboard has recently been expanded in a new peer-reviewed article – with references to our own work.
The Differences: How You Use Your Terabytes, How You Achieve Consensus
In spite of the commonalities, each one of these networks is quite different in how they achieve consensus and, notably, in what they use the storage for. Let us take a closer look at each of the protocols.
The vision of Filecoin is building a decentralised data storage marketplace. Therefore, storage is not used for consensus but for, precisely, storage. The consensus mechanism, in fact, is a variant of PoS. Storage, in Filecoin, is the use case.
Quite an opposite approach is undertaken by Chia. Chia leverages storage, but not as an application, as in terms of value proposition Chia is closer to your average protocol. Rather, storage is used for consensus, as an alternative to PoW that is significantly less energy intensive – and one that does not have the usual limitations associated with PoS. However, the data stored in Chia is not data useful in itself. Rather, a sort of arbitrary data structured (plots) are stored, and proving that a certain amount of hard drive space has been allocated for these plots and maintained over time is what will affect a validator’s level of influence on consensus. Chia data is thus useful for security rather than useful for retrieval of the data content. While not strictly PoW, one could argue that Chia relies on a form of “mining”, as by proving the storage of some space over time, a validator’s influence is attached to the amount of “work” that they can demonstrate.
Running a Filecoin node and participating in the network demands substantial storage and RAM, with a powerful GPU also recommended. Consequently, Filecoin nodes are typically operated in high-end data centres. In contrast, Chia requires less storage space and RAM, making it possible to run on a personal computer. Specifically, Filecoin nodes require at least 1 TB of storage and 128 GB of RAM for efficient sealing operations, often necessitating high-end data centre environments. On the other hand, Chia farming requires just 275 GB temporary storage and 4 GB of RAM, making it much more accessible for personal computers.
Finally, a unique species is found in Sia, and it is one that could be argued to be in the middle. In short, Sia’s consensus is literally PoW, however the use case of the blockchain is in itself decentralised data storage.
In summary:
- Filecoin is more closely related to PoS, but with storage as its application.
- Sia uses PoW, but with storage as its application.
- Chia uses a unique algorithm with some resemblance to mining, except influence over consensus is not attached to the amount of hashing but to the amount of storage space reserved for the blockchain.
Interpretation
The Filecoin Green team currently places Filecoin’s annualised energy consumption at 0.012 TWh (other sources place it at 0.013), whereas the Chia Power team places Chia’s at 0.13 TWh. No figures are currently available for Sia. The total storage of the networks is currently at 6.6 EB (exabytes) for Filecoin, 160 EB for Chia, and 0.00164 for Sia. In terms of sealing, Filecoin is currently adding 423 TB (terabytes) per day. Chia is adding 12 EB (exabytes), and this information is not known for Sia. There are differences in this sealing process, however: Filecoin’s is more energy-intensive because higher end hardware is used for plotting (GPUs), which is not the case for Chia and Sia. In addition, it is easier to understand the amount of plotting being done at Filecoin, as there are readily available figures on this. In contrast, this is not the case for Chia, which is understandable as the data stored for Chia is not used in itself. Moreover, Chia plotting figures can’t be approximated with the variation in its netspace, as this more closely reflects changes in data compression than plotting itself. For this reason, the Chia Power team actually assumes that plotting is a continuous expenditure, assuming an additional energy expenditure, spread over a lifespan of 2.5 years, for every byte.
Sia also has a hash rate of around 44 PH/s (petahashes per second). Although additional research is required to refine the numbers, it is clear that these networks consume significantly less energy than the major proof of work blockchains like Bitcoin by orders of magnitude, although slightly more than some of the largest proof of stake blockchains. Nevertheless, one should keep in mind that decentralised storage in the case of Filecoin and Sia constitutes a different use case to that of your typical blockchain for dApps.
Furthermore, any numbers that emerge from eventual research on the matter should be taken with care, as these blockchains are different from each other. Chia stores arbitrary data for security purposes, Filecoin saves data for decentralised storage, and Sia secures its data with proof of work. Hence the concept of “energy-efficiency” is relative to the utility assigned to each use case. Hardware used in every case is not identical, with hardware being much more lightweight for Chia, but initial plotting efforts being orders of magnitude more significant in terms of terabytes. Hence, Filecoin affords some initially high energy expenses, but the majority of the power usage comes from storage. Chia demonstrates much higher upfront electricity usage and needs to keep up this rate of new data plotting over time to contribute to security. Finally, while Sia energy expenses from data storage will be low after initial plotting, total energy usage will be higher in a more sustained manner as a continued proof of work is needed to maintain the network’s security.
In conclusion, understanding the energy consumption profiles of storage-based blockchains is crucial for evaluating their sustainability and environmental impact. Additional research is required to refine existing methodologies for storage-based networks. We are proud to say that, at the DLT Science Foundation, we are developing additional work in this direction. Stay tuned for more news!