PROMOTING RIGOUR IN BLOCKCHAIN SENERGY ENVIRONMENTAL FOOTPRINT RESEARCH A S YSTEMATIC LITERATURE REVIEW

2025-05-02 0 0 819.13KB 34 页 10玖币
侵权投诉
PROMOTING RIGOUR IN BLOCKCHAIN’SENERGY &
ENVIRONMENTAL FOOTPRINT RESEARCH: A SYSTEMATIC
LITERATURE REVIEW
A PREPRINT
Ashish Rajendra Sai
Department of Computer Science
Open Universiteit, Netherlands
&
SCET, University of California, Berkeley
ashish.sai@ou.nl
Harald Vranken
Department of Computer Science
Open Universiteit, Netherlands
&
Institute for Computing & Information Sciences
Radboud University, Netherlands
October 24, 2022
ABSTRACT
There is a growing interest in understanding the energy and environmental footprint of digital
currencies, specifically in cryptocurrencies such as Bitcoin and Ethereum. These cryptocurrencies are
operated by a geographically distributed network of computing nodes, making it hard to accurately
estimate their energy consumption. Existing studies, both in academia and industry, attempt to model
the cryptocurrencies energy consumption often based on a number of assumptions for instance about
the hardware in use or geographic distribution of the computing nodes. A number of these studies has
already been widely criticized for their design choices and subsequent over or under-estimation of the
energy use.
In this study, we evaluate the reliability of prior models and estimates by leveraging existing scientific
literature from fields cognizant of blockchain such as social energy sciences and information systems.
We first design a quality assessment framework based on existing research, we then conduct a
systematic literature review examining scientific and non-academic literature demonstrating common
issues and potential avenues of addressing these issues.
Our goal with this article is to to advance the field by promoting scientific rigor in studies focusing
on Blockchain’s energy footprint. To that end, we provide a novel set of codes of conduct for the
five most widely used research methodologies: quantitative energy modeling, literature reviews, data
analysis & statistics, case studies, and experiments. We envision that these codes of conduct would
assist in standardizing the design and assessment of studies focusing on blockchain-based systems’
energy and environmental footprint.
1 Introduction
All models are wrong, but some are useful.
George E. P. Box
This famous quote from the British statistician George E. P. Box highlights both the merit and limits of statistical
modeling. All models designed to represent real-world systems are inherently limited due to their reductive nature,
however, they may serve a useful purpose if designed and tested well and if their scope and assumptions are clearly
indicated. This is particularly true in the case of energy consumption models designed for sociotechnical systems [
1
].
Designing these models is a non-trivial task that requires a number of social, economic, and technical assumptions. The
intent behind many of these models is often to provide a useful insight in the form of an estimate of the the energy
arXiv:2210.11664v1 [cs.CY] 21 Oct 2022
APREPRINT - OCTOBER 24, 2022
requirement or environmental footprint, rather than an absolute measurement of energy consumed or carbon emission
by these systems.
Some of the early models from 2000s predicted the energy requirements of the internet and computers to a varying
degree of accuracy. With some early reports suggesting that all computers could consume up to 50% of U.S. electricity
in 2010 [
2
]. These claims have since been debunked through further research and empirical data [
3
]. This pattern of
inaccurate or misleading predictions and measurements regarding the energy consumption of a fast-growing information
technology is considered problematic as it may influence policymakers [
4
] and may feed misinformation to the general
public when picked up by popular media.
Decentralized digital assets are one such class of fast-growing information technology that have garnered significant
interest from both academia and industry due to its unique energy profile [
5
]. Bitcoin and other similar decentralized
digital assets often employ an energy-intensive consensus mechanism1known as Proof-of-Work.
By its design, the participants in the Proof-of-Work based digital assets are incentivized to spend considerable effort,
typically by executing compute-intensive or memory-intensive tasks, on a dynamically calibrated problem
2
. The first
participant to find and broadcast the solution to this problem within a dedicated time frame, is rewarded for their
participation in the form of newly minted cryptocurrencies. For example, on 1st June 2022, the reward to find the
solution or to mine one Bitcoin block was around 200K USD ([
7
]). This high reward induces an arms race to mine
the next block by spending more computational cycles on the problem. Each attempt to find a solution to the problem
incurs an energy cost in the form of electricity spent to power the device that solves the problem.
Similar to the early days of the internet and computers, we have seen numerous attempts at measuring the electricity
consumption of decentralized digital assets such as Bitcoin [
8
]. It has been a frequent sight to see news headlines
indicating the colossal energy and environmental footprint of Bitcoin. Many of the non-academic literature and (highly
rated) academic sources used in these news headlines have been criticized for inaccuracy or misleading interpretations
([9, 10, 11]).
While we acknowledge that it is worthwhile to explore the energy and environmental footprint of cryptocurrencies such
as Bitcoin, we stress that this should be done with utmost care to avoid inaccurate analysis and unjustified assumptions
that may lead to sensational news headlines. For instance, the article published by [
12
] suggested that Bitcoin alone
could push global warming above 2 degrees Celcius as soon as 2033. This article has been widely criticized for provably
inaccurate underlying assumptions such as participants using unprofitable hardware ([10, 11, 13, 14]).
As it is inherent with energy modeling, each of these models rely on several assumptions to provide an estimate,
thus their accuracy is subject to the validity of their underlying assumptions. The scientific expectation is that these
assumptions are not only mentioned explicitly but also be backed by verifiable, preferably empirical evidence or
justification ([15]).
Unfortunately as seen in the case of [
12
], it is not always the case. Further research into the reliability of these studies
by [
11
,
8
] has suggested that these issues are not isolated to one particular study. However as they both have only
focused on a small set of models, it is difficult to generalize the results to the whole field.
Our study attempts to overcome this limitation by conducting a systematic literature review of both scientific and
non-academic literature focusing on the energy and environmental footprint of cryptocurrencies. We assess the quality
of the shortlisted literature against the guidelines put forth by Lei et al. (2021) and Sovacool et al. (2018) [15].
We iteratively refine our quality assessment framework to account for domain-specific variations
3
. Thus, in this work,
we present the first in-depth analysis of scientific rigor of blockchain energy and environmental models in order to
assess the following question:
Are the existing energy and environmental footprint models and resulting estimates for blockchain-based systems
trustworthy?
It is important to note that the purpose of our article is not to discuss whether or to what extent specific studies are
flawed but to provide tools to transparently discuss the rigor of these studies while allowing for improvements in the
1
In distributed computing systems such as Peer-to-Peer network-based cryptocurrencies, a consensus mechanism is employed to
achieve an agreement on a single view of the data such as a ledger of transactions. We refer the reader to [
6
], for further information
on consensus mechanisms in blockchain-based systems.
2
In Bitcoin like Proof-of-Work based cryptocurrencies, the participants are tasked with the problem to find a block hash value
below a set threshold. The difficulty of this problem is periodically changed to maintain the system property of 10 minutes time
difference between two blocks of transactions.
3
This is particularly important for the guidelines provided by Sovacool et al. (2018), as these guidelines are not specific to the
blockchain domain.
2
APREPRINT - OCTOBER 24, 2022
design and prediction of these models. We support and encourage the work done in Blockchain energy sciences over
the last few years and intend to expand on it through this review.
To study the reliability of energy models, we first coded and analyzed relevant scientific and non-academic literature.
We review the literature published from 2008 on, i.e. post the introduction of Bitcoin’s white paper [
16
]. This is done
by following the guidelines proposed by [
17
]. As suggested by Kitchenham et al. (2007), our review is broken down
into five steps: Search, Selection, Quality Assessment, Data Extraction, and Analysis. This review results in an article
pool of 128 studies. These articles are then assessed for their scientific rigor by using the quality assessment framework
based on the guidelines of [15] and [8].
Following the assessment of the scientific rigor, we consolidate our findings in the form of commonly occurring issues
in different types of studies. We also document potential avenues of addressing these known issues. This subsequently
leads to the development of novel code of practices to promote scientific rigor in blockchain energy studies.
We believe that this study assists the reader in understanding the reliability of the current energy and environmental
studies in a blockchain context. This article also assists researchers and developers in designing or refining their existing
models through adherence to the code of practice. The paper makes the following contributions:
•
We systematically review the existing literature to document common issues with energy and environmental
impact studies for Blockchain-based systems (Section 3).
•
We develop a novel quality assessment framework for Blockchain-specific studies that can assist in under-
standing the scientific rigor of the energy or environmental impact model (Section 3).
•
We identify research gaps specifically with regards to the lack of non-Bitcoin-specific investigations in
academic literature. We also report on the lack of empirical data for these models (Section 4.1).
•
We manifest the findings of our review in a set of best practices that can assist in designing or improving
existing models (Section 5). 6).
2 Background
Cryptocurrencies that use an energy-intensive consensus mechanism cause two prime concerns from an environmental
perspective: the electricity consumption and the carbon emission associated with the energy consumption
4
. In this
section, we provide an overview of how the energy and environmental footprints of these cryptocurrencies are usually
measured.
2.1 Energy Consumption
As alluded to in the introduction section, measuring the energy consumption of a geographically distributed network
is a non-trivial task. This problem is compounded when considering decentralized systems as it is difficult to find a
centralized source of information about the network’s physical composition [
19
]. There are two main ways of estimating
the energy consumption of a blockchain-based system depending on the availability of reliable data on the computing
network: bottom-up and top-down.
2.1.1 Bottom-Up
A distributed computing network is made up of computational devices that consume a certain amount of electricity per
unit of work
5
. Each of these computational devices can have different performance and energy efficiency profiles. For
example, a network could be made up of 100 Raspberry Pi
6
devices generating X unit of work in a single unit of time or
it could be made up of 2 consumer-level personal computers generating the same amount of work in the same time
resolution.
One of the early attempts at using a bottom-up approach for modeling electricity consumption of Bitcoin was made by
[
20
]. In his analysis, [
20
] outlined prominent modern bitcoin mining hardware that typically employ application-specific
integrated circuits (ASICs) designed for bitcoin mining. For example, an Antminer S9 system released in 2017 could
4
There are other environmental impacts associated with cryptocurrency operations such as E-Waste generation [
18
], we briefly
touch on this in a subsequent section however our focus in this article is primarily on energy consumption and carbon emissions
5
In Proof-of-Work based cryptocurrencies, the work is often performing hashing operations to find a nonce value such that the
resulting hash value is lower than the target.
6See www.raspberrypi.com.
3
APREPRINT - OCTOBER 24, 2022
Figure 1: Calculating Energy Consumption using Bottom Up Approach
perform 13 TH/s whereas a consumer CPU such as Intel i7 (2021) can only perform 2.5 KH/s while being more power
efficient per unit of hash calculation.
If we are aware of the exact hardware used in the network, including the hardware distribution (how many units of
each type of device are on the network), we can use this information to calculate the total energy consumed by all the
constituting computing devices.
This calculation requires accurate values of each device’s computing power and energy efficiency. This in itself can be
problematic in a real-world scenario, as most of the information about power and energy efficiency is obtained through
data sheets provided by manufacturers. These data sheets in most cases are not verified by an independent auditor.
Furthermore, tuning operational parameters like clock frequency and supply voltage may even lead to different numbers
in practice. For an accurate measurement, it is also important to know the uptime for each device and the actual work
done during this uptime.
This gives us a partial understanding of the network’s energy consumption as this calculation does not consider
operational electricity consumption for devices other than the computing device such as the networking or cooling
infrastructure. These operational expenses are often considered in the form of a fractional value known as Power Usage
Effectiveness (PUE) [21].
Once we know the energy consumption of each device in the network and the associated PUE value, we can calculate
the total energy consumed as follows 7:
T=Xε(i)∗P UE(i)(1)
Where
T
is the total energy consumption,
ε
is the energy consumption of each constituting computing device and PUE
is the additional operational electricity requirement. This calculation is also visually illustrated in Figure 1.
2.1.2 Top-Down
Bitcoin and other cryptocurrencies are often described as decentralized systems. Decentralization is a crucial component
for the network on different levels of operations such as applications (decentralized exchanges), protocol (consensus
mechanism), and network (distributed peer-to-peer network) [
19
]. This decentralized nature of the network makes it
difficult for researchers to collect empirical data on the location and hardware of the consensus participants8.
Due to the lack of reliable empirical information on the network participants, a large number of energy and environmental
models are based on a top-down approach [
22
]. In a top-down modeling approach, the model relies on high-level
technical, economical or social variables.
For instance, [
23
]’s top-down model is based on the total computing power also known as hash rate. In [
24
]’s model,
the author builds an economic model based on the economic rationality of the miner.
In this subsection, we provide an abstract overview of how a Top-Down model conceptually works, however, we refer
the reader to [8, 22, 23] for an in-depth discussion of top-down modeling.
7
It is worth noting that this equation is for illustrative purpose only as in the real model, the authors might account for additional
factors such as economic of operators.
8In Proof-of-Work based systems, these consensus participants are also known as miners.
4
APREPRINT - OCTOBER 24, 2022
Figure 2: Calculating Energy Consumption using a Top Down approach
For Bitcoin, we can calculate the total hash rate of the network by using the difficulty of mining [
25
]. First of all, an
estimate on the hash rate of the network is required. This can be derived from a simple statistical model that considers
the difficulty parameter and the time it takes on average to mine a block, which is 10 minutes for Bitcoin, as is applied
by [
25
]. A more refined model may use empirical data on the exact amount of time it takes to mine blocks. For instance,
for Bitcoin the difficulty parameter is adjusted every 2,016 blocks, and hence some drift may occur in between.
The total hash rate of the network is composed by the combined hash rates of a number of different hardware in
use, each with a specific energy and performance profile. A number of different combinations and permutations of
available hardware can generate the required hash rate. Different models make different assumptions in order to
get to the total hash rate. For example, some models assume that the network is made up of only the most efficient
commercially available hardware while others consider a pool of hardware with different distributions. We can represent
this calculation as follows 9:
H=Xρ(2)
Here
H
is the hashing power of the network composed of all the individual hashing power (
ρ
) of the hardware used in
the network. Once a pool of hardware is decided upon, we can similarly calculate the energy consumption to that of
bottom-up10:
T=Xε(i)∗P UE(i)(3)
We have also illustrated this process in Figure 2.
2.2 Environmental Impact Measurement
The scope of the environmental impact of information technologies can be very broad ranging from the direct impact
caused by E-waste and through the consumption of electricity generated by non-renewable carbon intensive operations
such as coal-based power plants [
26
]. Through our literature review, we report that most of the studies in the blockchain
context focus on the carbon emission associated with the electricity consumption of the network. However, it is worth
noting that there are a few studies that look at other aspects of the environmental impact of cryptocurrencies such as
E-waste generation [18] and scope 2 and 3 carbon emissions [27]. In this subsection, we focus on carbon emissions.
The carbon emission calculation consists of a five-step process as outlined below11:
1.
Calculating the total energy consumed by the network: This can be determined by either Bottom-Up or
Top-Down approaches as discussed above.
2.
Determining the geographic location of the devices in the network: In addition to understanding the pool of
hardware and their respective share of the total network, we also need to know their geographic location.
9
There are often many different combinations of devices possible here with different
ρ
. For instance, a small network can be
made up of a large number of inefficient devices or a small number of highly efficient devices.
10The εis only for hardware that contribute to Habove.
11It is important to note that these steps are only indicative of the process, individual studies may differ in their approach.
5
摘要：

PROMOTINGRIGOURINBLOCKCHAIN'SENERGY&ENVIRONMENTALFOOTPRINTRESEARCH:ASYSTEMATICLITERATUREREVIEWAPREPRINTAshishRajendraSaiDepartmentofComputerScienceOpenUniversiteit,Netherlands&SCET,UniversityofCalifornia,Berkeleyashish.sai@ou.nlHaraldVrankenDepartmentofComputerScienceOpenUniversiteit,Netherlands&Ins...

展开>> 收起<<
PROMOTING RIGOUR IN BLOCKCHAIN SENERGY ENVIRONMENTAL FOOTPRINT RESEARCH A S YSTEMATIC LITERATURE REVIEW.pdf

共34页,预览5页

还剩页未读， 继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

相关推荐

更多
立即下载
分类：图书资源 价格：10玖币 属性：34 页 大小：819.13KB 格式：PDF 时间：2025-05-02

开通VIP享超值会员特权

  • 多端同步记录
  • 高速下载文档
  • 免费文档工具
  • 分享文档赚钱
  • 每日登录抽奖
  • 优质衍生服务

作者详情

相关内容

更多

热门标签

人际关系 动力学连接体 力的合成 配电装置 高考理综 全宋诗作者索引 公务员考试
/ 34
评分 收藏
分享
立即下载
客服
关注