PaperlayerView exact revision on arXiv

Source-linked AI summary

Blockchain Consensus Algorithms: A Survey

Md Sadek Ferdous, Mohammad Jabed Morshed Chowdhury, Mohammad A. Hoque, Alan Colman

arXiv:2001.07091v2cs.DC

TL;DR

Blockchain consensus algorithms shape system behavior, and varied application requirements make their applicability important to examine. The survey applies a taxonomy to consensus algorithms and analyzes cryptocurrencies; it finds PoW remains dominant despite disadvantages and limited adoption of PoS and DPoS.

  • Problem

    Different blockchain applications have distinct requirements, motivating examination of whether existing consensus algorithms apply in newer settings and development of novel mechanisms.

    Source: §1 · p. 1

  • Method

    The paper introduces a taxonomy of consensus properties, groups algorithms as incentivised or non-incentivised, and uses the taxonomy to analyze them.

    Source: §9 · p. 33

  • Results

    PoW accounts for 57% of the top 100 cryptocurrencies’ consensus algorithms, compared with 11% for DPoS and 6% for PoS, while PoW has reduced throughput relative to PoS and DPoS.

    Source: §8 · pp. 29–30

  • Takeaways & Limitations

    PoS and DPoS aim to address PoW limitations, but their adoption remains limited compared with PoW.

    Source: §9 · p. 34

    PoW’s mining centralization concern may also leave PoW currencies vulnerable to collusion attacks.

    Source: §8 · p. 31

Abstract

from arXiv · show

In recent years, blockchain technology has received unparalleled attention from academia, industry, and governments all around the world. It is considered a technological breakthrough anticipated to disrupt several application domains. This has resulted in a plethora of blockchain systems for various purposes. However, many of these blockchain systems suffer from serious shortcomings related to their performance and security, which need to be addressed before any wide-scale adoption can be achieved. A crucial component of any blockchain system is its underlying consensus algorithm, which in many ways, determines its performance and security. Therefore, to address the limitations of different blockchain systems, several existing as well novel consensus algorithms have been introduced. A systematic analysis of these algorithms will help to understand how and why any particular blockchain performs the way it functions. However, the existing studies of consensus algorithms are not comprehensive. Those studies have incomplete discussions on the properties of the algorithms and fail to analyse several major blockchain consensus algorithms in terms of their scopes. This article fills this gap by analysing a wide range of consensus algorithms using a comprehensive taxonomy of properties and by examining the implications of different issues still prevalent in consensus algorithms in detail. The result of the analysis is presented in tabular formats, which provides a visual illustration of these algorithms in a meaningful way. We have also analysed more than hundred top crypto-currencies belonging to different categories of consensus algorithms to understand their properties and to implicate different trends in these crypto-currencies. Finally, we have presented a decision tree of algorithms to be used as a tool to test the suitability of consensus algorithms under different criteria.

1 INTRODUCTION

Blockchain systems depend on consensus algorithms for agreement and performance, but differing application requirements and gaps in prior analyses motivate a systematic comparison. This survey organizes algorithms and related blockchain systems, examines their issues, and offers a decision tree for assessing suitability.

1 INTRODUCTION

  • The survey identifies prior analyses as incomplete in their property coverage, algorithm and mechanism scope, and practical links between blockchain systems and consensus algorithms.It calls for a concise but comprehensive framework grounded in existing algorithms and systems.Source: §1 · p. 2
  • A novel taxonomy groups consensus algorithms as incentivised or non-incentivised, further subdivides them, and applies shared properties to each sub-category.The sub-category analyses are also presented in tables for visual comparison.Source: §1 · p. 2
  • The survey examines major issues in each consensus-algorithm category and analyses their implications.Source: §1 · p. 2
  • More than 100 cryptocurrencies across consensus categories are analysed for properties and trends, linking systems with their underlying algorithms.Source: §1 · p. 2
  • A decision tree is presented to test consensus-algorithm suitability under selected criteria.Source: §1 · p. 2

2 BACKGROUND: DISTRIBUTED CONSENSUS

Distributed consensus coordinates a shared state across nodes by ordering common inputs, while protocol behavior depends on assumptions about failures and network timing. The section outlines crash- and Byzantine-tolerant approaches and the limits imposed by asynchronous settings and partitions.

2 BACKGROUND: DISTRIBUTED CONSENSUS

  • State Machine Replication makes consistent replicated state possible when nodes receive the same inputs in the same order, a task supported by consensus protocols that disseminate inputs and ensure atomic broadcast.A deterministic state machine processes ordered inputs to evolve its state.Source: §2 · p. 2
  • Protocols rely on assumptions about networking and failure behavior, including synchronous, asynchronous, and eventually synchronous networks and crash or Byzantine failures.Source: §2 · pp. 2–3
  • In asynchronous networks, FLP rules out satisfying consistency, availability, and fault tolerance together; a partition can likewise prevent a replicated datastore from achieving both consistency and availability.Source: §2 · p. 3
  • Crash-tolerant protocols can function when t < n/2, while PBFT tolerates f < n/3 Byzantine nodes under an eventual-synchrony assumption.Both classes use leaders and epochs or views to order messages; crash-tolerant protocols elect a new leader after a crash.Source: §2 · p. 3

3 BACKGROUND: BLOCKCHAIN

The section introduces blockchain structure and operation, from Bitcoin’s block creation and fork resolution to blockchain properties, deployment types, and a four-layer model. It distinguishes blockchain from the broader category of distributed ledgers and assigns different functions to each layer.

3 BACKGROUND: BLOCKCHAIN

  • A blockchain is a distributed ledger of linked blocks containing transactions, whereas a distributed ledger can be any ledger distributed across a network.Source: §3 · pp. 3–4

3.1 Bitcoin

  • Bitcoin miners collect transactions, compete to solve a target-based cryptographic puzzle, and propagate valid blocks; when forks occur, miners select the branch with the greatest cumulative computational effort.Only miners on the surviving main branch receive rewards; Bitcoin adjusts difficulty every 2016 blocks to target an average block interval of 10 minutes.Source: §3.1 · p. 4

3.2 Properties of blockchain

  • Blockchain properties include consensus without a trusted third party, practical immutability, persistent distributed data, provenance, distributed control, and transparency.Source: §3.2 · p. 5

3.3 Blockchain type

  • Public blockchains allow anyone to participate and expose state and transactions, while private blockchains restrict participation to authorized, trusted entities and can preserve chain-data privacy.Source: §3.3 · p. 5

3.4 Blockchain Layers

  • The proposed four layers are network, consensus, application, and meta-application, grouping communication, block ordering, system semantics, and semantic overlays respectively.The authors group mining with consensus because mining can be integral to consensus, while some systems have no mining algorithm.Source: §3.4 · pp. 5–6

4 CONSENSUS TAXONOMY & PROPERTIES

The section classifies consensus as incentivised or non-incentivised, then sets out property groups for comparing algorithms and their blockchain systems. Its taxonomy covers structure, block and reward, security, and performance.

4 CONSENSUS TAXONOMY & PROPERTIES

  • Incentivised consensus rewards block creators in public blockchains, while non-incentivised consensus is used in private systems with authorised participating nodes.The reward is described as an incentive to follow the protocol; without rewards, participating nodes are considered trusted.Source: §4 · p. 6

4.1 Consensus properties

  • The evaluation taxonomy groups consensus properties into structural, block and reward, security, and performance categories.The properties are compiled from prior research to provide criteria for comparing algorithms with different characteristics and purposes.Source: §4.1 · p. 6

4.1.1 Structural properties

  • Structural properties describe how nodes participate, including single or multiple committees and open or closed committee membership.The taxonomy also distinguishes implicit from explicit committee formation and includes node types and selection mechanisms.Source: §4.1.1 · pp. 6–7
  • Committee configuration may be static or dynamic, while selection mechanisms include lottery, coin age, or voting.Dynamic committees can change membership over an epoch or, in some algorithms, allow members to join or leave at any time.Source: §4.1.1 · pp. 6–7

4.1.2 Block & reward properties

  • Block and reward properties used to distinguish cryptocurrencies include genesis date, block reward, total supply, formula, and block creation time.Most of these properties, except genesis date, are described as having a direct or indirect impact on how consensus is achieved.Source: §4.1.2 · p. 7, Figure 5 · p. 7

4.1.3 Security properties

  • Security properties include authentication, non-repudiation, censorship resistance, and attack vectors such as adversary tolerance and Sybil attacks.The taxonomy also includes DoS resistance; some attack vectors are specific to particular classes of consensus algorithms.Source: §4.1.3 · pp. 7–8

4.1.4 Performance properties

  • Performance properties measure fault tolerance, throughput, scalability, latency or finality, and energy consumption.Latency or finality is the time from transaction proposal until consensus is reached.Source: §4.1.4 · p. 8

5 INCENTIVISED CONSENSUS: POW & POS

The section groups incentivised consensus into Proof of Work, Proof of Stake, and hybrid algorithms; it discusses PoW and PoS here and presents hybrid algorithms in Section 6.

5 INCENTIVISED CONSENSUS: POW & POS

  • Incentivised consensus is grouped into Proof of Work, Proof of Stake, and hybrid algorithms, with hybrid algorithms presented in Section 6.This section discusses PoW and PoS in Sections 5.1 and 5.2, respectively.Source: §5 · p. 8

5.1 Proof of Work (PoW)

This section surveys PoW’s proof mechanisms and compute-bound, memory-bound, and chained variants, including approaches to mining participation and ASIC resistance, and describes its role in distributed consensus. It also examines PoW’s open participation, security and performance properties, and limitations, including energy use, mining centralization, economic pressures, and the absence of penalties.

5.1 Proof of Work (PoW)

  • PoW makes proof generation resource-intensive while verification requires significantly fewer resources, and blockchain miners repeat attempts until a proof meets a difficulty threshold.The resource asymmetry is described as a deterrent to system abuse; each attempt is a round of the task.Source: §5.1 · p. 8

5.1.1 Compute-bound PoW

  • Compute-bound PoW uses CPU/GPU computation that can be optimized in ASICs, while Nakamoto consensus adapts Hashcash to find blocks below a difficulty target and resolve forks.Hashcash requires repeated SHA-1 attempts for a hash with at least 20 leading zero bits; Nakamoto miners use SHA-256 or SHA-256d. ASIC optimization is criticized for concentrating mining among fewer nodes.Source: §5.1.1 · p. 9

5.1.2 Memory-bound PoW

  • Memory-bound PoWs constrain performance through memory access and capacity; Scrypt relies on substantial memory, while Equihash uses a steep time-memory trade-off to support ASIC resistance.The section presents memory-bound PoW as a way to limit ASIC advantages and reduce mining concentration, while noting that rising coin prices have nevertheless encouraged ASIC hardware for some Scrypt currencies.Source: §5.1.2 · pp. 9–10
  • Ethash combines Dagger and Hashimoto around a large, epoch-recomputed DAG whose memory accesses bottleneck mining, making comparable ASIC performance require substantial memory.The dataset is recomputed every 30,000 blocks, approximately five days; the section also notes that Dagger was dropped as an Ethereum PoW candidate after a shared-memory acceleration vulnerability.Source: §5.1.2 · pp. 10–12

5.1.3 Chained PoW

  • Chained PoW combines hashing functions to reduce reliance on a single hash and continue functioning until all chained functions are broken; X11 uses eleven hashes and is reported as more energy-efficient than Scrypt.The X11 comparison reports GPUs using approximately 30% less wattage and remaining 30–50% cooler than with Scrypt, yet economic incentives led to ASIC mining; Quark ASIC hardware also appeared despite its CPU-mining restrictions.Source: §5.1.3 · p. 12

5.1.4 PoW Limitations

  • The section presents Nakamoto PoW as a breakthrough that laid groundwork for later PoW algorithms and currencies, then signals a discussion of its significant limitations.The supplied passage introduces those limitations but does not specify them.Source: §5.1.4 · pp. 13–14
  • 40 TWh for Bitcoin and 10 TWh for Ethereum are reported annual energy consumption levels; Bitcoin’s use could power 6,770,506 American households, raising sustainability concerns.The household comparison is presented relative to country electricity use.Source: §5.1.4 · p. 14, Figure 10 · p. 14
  • Five mining pools control 75% of Bitcoin’s hashrate, as rising mining costs and economies of scale centralize block creation and raise concern about collusive 51% attacks.The hashrate distribution is shown in Figure 11.Source: §5.1.4 · pp. 14–15, Figure 11 · p. 15
  • When Bitcoin’s block reward nears zero, miners may compete for transaction fees, driving fees down and potentially pushing miners out and destabilizing the network.The passage describes this as a tragedy-of-the-commons risk that could lead toward 51% attacks.Source: §5.1.4 · p. 15
  • PoW rewards behaving miners but does not penalize misbehavior, enabling selfish mining, transaction or block non-forwarding, and repeated spawn-camping DoS attacks.The passage proposes that penalties could disincentivize malicious behavior.Source: §5.1.4 · p. 15

5.1.5 Analysis

  • PoW uses miners to create blocks by randomized lottery and clients to validate blocks and transact; its single open committee lets miners dynamically join or leave.The committee represents the set of miners.Source: §5.1.5 · p. 15
  • PoW algorithms require no participant authentication, support transaction-signature non-repudiation and censorship resistance, and assume adversaries control less than 50% of network hashing power.Substantial mining-hardware investment is described as deterring Sybil and DoS attacks.Source: §5.1.5 · p. 15
  • PoW has low throughput, poor scalability, often slow finality, and high energy use; its stated fault tolerance is 2f + 1, requiring more than 50% of nodes to function correctly.The energy use is discussed in Section 5.1.4.Source: §5.1.5 · p. 15

5.2 Proof of Stake

This section explains how Proof of Stake uses staked coins to select block creators and surveys chained, BFT, and delegated variants, including their mechanisms and limitations. It compares PoS approaches and DPoS designs in terms of committee structures, security characteristics, performance, and algorithm properties.

5.2 Proof of Stake

  • PoS participants escrow coins as stake to join block creation, with misbehavior risking loss of stake; obtaining and fairly distributing initial coins is a key bootstrap problem.The proposed bootstrap approaches are pre-mining followed by an IPO or ICO, or starting with PoW and gradually transitioning to PoS; minters may be rewarded through transaction fees or interest.Source: §5.2 · pp. 15–16
  • DPoS uses reputation or other mechanisms to let network users elect delegates, who alone propose blocks; its small validator set can enable fast consensus finality.A round leader is selected from the delegates, and misbehavior can lead to penalties and removal from the validator set.Source: §5.2 · p. 17

5.2.1 Chained PoS

  • Peercoin combines PoW and PoS blocks, using coin-age-weighted staking rewards and difficulty adjusted to target one block every 10 minutes.Its PoS blocks reward minters approximately 1% per annum based on staked coin-age; the survey notes chained PoS generally favors availability over consistency during network partitions.Source: §5.2.1 · p. 17, §5.2 · p. 17
  • Casper FFG overlays PoS checkpoint finalization on Ethereum’s PoW chain, but the survey says it was not yet implemented and its real-world performance remained unknown.Every 50 blocks, validators vote across two successive epochs: a checkpoint needs more than two-thirds support to be justified, and support for its child finalizes it; the proposal favors availability over consistency and is vulnerable to cartel formation.Source: §5.2.1 · p. 18

5.2.2 BFT PoS

  • Tendermint commits blocks only after supermajority voting and locking, guaranteeing no conflicting commits when fewer than one-third of validators are Byzantine.Its safety comes with an availability cost: if more than one-third of validators are offline or partitioned, the system stops functioning and requires out-of-protocol steps.Source: §5.2.2 · pp. 18–19
  • CTFG favors availability over consistency and can theoretically accommodate more dynamic validators than Tendermint, but remains a proposal whose real-world performance is untested.It uses a formal Correction by Construction model with GHOST as its fork-choice rule; validators bet on blocks and are rewarded or penalized according to whether their choice enters the main chain.Source: §5.2.2 · p. 19
  • Ouroboros is described as a provably secure PoS algorithm used by Cardano, which supports smart contracts and decentralized applications without PoW.Source: §5.2.2 · p. 19
  • Ouroboros elects a slot leader for each block using a random seed and coin selection; it is proven secure against almost all PoS attacks except cartel formation, but deployment performance remains unknown.Stakeholders hold Ada, and electors jointly generate the seed through multiparty computation; greater coin ownership raises a stakeholder’s chance of selection. Transaction-fee rewards were still being finalised.Source: §5.2.2 · pp. 19–20

5.2.3 DPoS

  • EOS uses token-holder votes to select 21 block producers, with selection frequency proportional to votes received; producers create blocks in rounds and require more than two-thirds agreement for confirmation.Each round comprises 21 blocks, with each selected producer getting a chance to produce one. A block is treated as final after the required agreement.Source: §5.2.3 · p. 20
  • Tezos lets users delegate XTZ to help bakers meet participation thresholds, while Lisk uses 101 vote-selected delegates and supports application sidechains alongside its mainchain.Tezos has no predefined number of bakers and returns a portion of a baker’s block reward to delegators. Lisk’s 101 delegates are selected before each round of 101 blocks.Source: §5.2.3 · p. 20

5.2.4 Limitations of PoS

  • Small validator sets can make collusive 51% attacks easier, while reliance on coin wealth gives large holders greater influence and cost-free voting across forks enables nothing-at-stake attacks.The survey says penalties can address nothing-at-stake misbehaviour.Source: §5.2.4 · p. 21
  • Short-range bribery can make a double-spend fork profitable, while long-range attacks build an alternative chain from early blocks; penalties and checkpointing are proposed countermeasures, respectively.In the short-range scenario, an attacker forks after a seller delivers goods and bribes minters to extend the fork; checkpointing makes the chain immutable up to a specified point.Source: §5.2.4 · p. 21
  • Uncapped coin-age enables double-spend attacks, weak randomness enables stake grinding, and stake cartels may monopolize PoS or launch 51% attacks.The survey proposes capping coin-age and adding a secure randomness source to address the first two attack vectors.Source: §5.2.4 · p. 21

5.2.5 Analysis

  • Most PoS algorithms use open, dynamically configured single committees and voting; chained algorithms use multiple dynamic committees and lotteries, while Tendermint has a closed static committee and is the sole algorithm requiring authentication.Chained systems and most other PoS algorithms support adversary tolerance of 3f + 1, while BFT Ouroboros supports 2f + 1.Source: §5.2.5 · pp. 21–22
  • BFT and DPoS algorithms have high throughput, low latency, high scalability, and negligible energy use; chained algorithms compare less favorably on throughput, scalability, and latency.The analysis also reports that CTFG, Tendermint, and Ouroboros mitigate the additional attack vectors examined; Casper FFG and DPoS do not defend against cartel formation, while Peercoin mitigates cartel formation but suffers from the other listed vectors.Source: §5.2.5 · p. 22

6 INCENTIVISED CONSENSUS: BEYOND POW AND POS

The section groups incentivised algorithms beyond conventional PoW and PoS into hybrid designs and mechanisms that use neither, then compares their properties. Its analysis finds important differences in structure, security, attack resistance, and performance.

6.1 Hybrid Consensus

  • Hybrid PoR combines PoS with Proof-of-BOINC research contributions, but PoS dominates consensus and Proof-of-BOINC rewards researchers for sharing computing resources.Gridcoin researchers submit BOINC work and are credited based on recommendations compared with another researcher; the passage says PoR security resembles traditional PoS.Source: §6.1 · p. 22
  • PoB makes coin burning a condition for mining selection, with greater burning associated with a greater chance of selecting the next block miner.Slimcoin combines PoB with PoW and PoS; PoW generates the initial supply, then PoB selects minters for PoS, and its security is described as mostly similar to Peercoin's.Source: §6.1 · pp. 23–24
  • PoSV adds stake velocity to PoS to encourage minters to stake actively, addressing the participation gap that coin-age-based PoS can leave.Its nonlinear coin-ageing function rewards stake gained faster in the first days and weeks; minters staking every two weeks or less may earn up to 20% more than nonparticipants.Source: §6.1 · p. 24

6.2 N-POS/POW

  • PoC uses authenticated Certified Validating Nodes, community and technical eligibility rules, and round-robin block creation with voting to select the next creator.Each CVN independently selects a candidate; the candidate proceeds after receiving selections from more than 50% of CVNs, then propagates a signed block for validation.Source: §6.2 · p. 24
  • PoI selects eligible block creators using an importance score that combines vested currency with transaction activity, rather than stake alone.The score uses net transfers over 30 days, vesting, and transaction-cluster relationships; accounts need at least 100 XEM transferred during that period to be considered.Source: §6.2 · pp. 24–25

6.3 Analysis

  • PoB, PoSV, and PoI use open, dynamic single committees; PoR uses a dynamic flat multi-committee, while PoC uses an implicit, open, dynamic single committee and voting.The comparison also reports PoR and PoB as having flat topology, while PoSV and PoI use probabilistic lotteries.Source: §6.3 · p. 25
  • The compared algorithms are reported to support non-repudiation, Sybil protection, DoS resistance, and high censorship resistance, but PoC and PoI lack reported adversary-tolerance information.The passage also says PoB, PoSV, and PoI entities need not be authenticated, whereas PoC entities and PoR researchers do.Source: §6.3 · p. 25
  • PoR, PoSV, and PoI have high throughput, while PoC is low-throughput and PoB medium-throughput; all but PoB are described as low-latency and low-energy.PoB is described as low-scalability, low-latency, and medium-energy; fault-tolerance information is unavailable for PoC and PoI.Source: §6.3 · p. 25

7 NON-INCENTIVISED CONSENSUS

The section surveys non-incentivised consensus in private blockchain systems, describing several Hyperledger projects and comparing their structures, security, and performance. Its analysis contrasts scalable lottery-based PoET with voting-based systems that have lower latency but limited validator scalability.

7.1 Hyperledger Fabric

  • Fabric separates transaction endorsement from ordering: endorsers validate and endorse transactions, while orderers create blocks and extend the ledger.The Membership Service Provider authenticates participants, and identity-based policies control permitted actions; Fabric supports multiple ledgers and pluggable consensus.Source: §7.1 · pp. 25–26

7.2 Hyperledger Sawtooth

  • Sawtooth's PoET uses Intel SGX to assign validators wait times, selecting the shortest-wait validator to propose a block.Other validators verify the trusted wait-time process and block; the passage says avoiding computationally intensive puzzles enables scalability and use on permissioned or public ledgers.Source: §7.2 · p. 27

7.3 Hyperledger Burrow

  • Burrow combines a permissioned EVM with Tendermint and known validators; the paper says this supports smart contracts and faster scaling than Ethereum while preserving transaction privacy.The stated Byzantine tolerance is one-third of nodes exhibiting Byzantine behavior, including malicious action or failures due to network or system issues.Source: §7.3 · p. 27

7.4 Hyperledger Iroha

  • Iroha emphasizes mobile-friendly, simpler infrastructure and fine-grained permissions, with YAC handling consensus and distinct validation of transaction format, signatures, and business rules.Its architecture includes clients, peers that maintain ledgers and participate in consensus, ordering-service nodes, and components for networking, storage, simulation, and synchronization.Source: §7.4 · pp. 27–28

7.5 Hyperledger Indy

  • Indy is built for self-sovereign identity and uses RBFT, where a primary orders transactions through three-phase commit and backup instances monitor its performance.Its Plenum ledger maintains ordered transactions, while Indy-SDK provides APIs and tools for software to interact with the ledger.Source: §7.5 · pp. 28–29

7.6 Analysis

  • The systems use closed, explicitly formed single committees; only YAC is described as dynamically configured, while the others use static configuration.Voting is used by Tendermint Burrow, YAC, and RBFT; PoET uses a lottery, and Fabric uses ordering services, with SBFT described as a future possibility.Source: §7.6 · p. 29
  • All surveyed systems support non-repudiation and have low censorship resistance; all except PoET are reported to tolerate 3f + 1 adversarial nodes.The text attributes low censorship resistance to participants having known identities, which also supports Sybil protection and DoS resistance; PoET's adversarial tolerance is given as Θ[220].Source: §7.6 · p. 29
  • PoET can scale to many validators, but that increases transaction latency; voting-based systems provide low latency but do not scale to large validator sets.All compared systems are described as offering good throughput and requiring no significant energy; Fabric, YAC, and RBFT provide 2f + 1 fault tolerance, while information for PoET and Tendermint Burrow is unspecified.Source: §7.6 · p. 29

8 DISCUSSION

The discussion finds that PoW remains dominant in the sampled top 100 cryptocurrencies despite its energy and throughput drawbacks, while adoption and market-cap patterns reflect Bitcoin's weight and other factors. It closes with a decision tree for matching algorithms to criteria, while noting that criteria and resulting recommendations may vary by scenario.

8 DISCUSSION

  • PoW accounts for 57% of consensus algorithms in the top 100 cryptocurrencies, compared with 11% for DPoS and 6% for PoS.The analysis uses the top 100 CoinMarketCap cryptocurrencies as of 18 July 2019; all other algorithms make up the remaining 26%.Source: §8 · p. 29, Figure 15 · p. 31
  • PoW remains the most-used algorithm among currencies created in 2017, 2018, and 2019: 11, 19, and 4 currencies, versus 4, 2, and 2 for PoS and DPoS combined.The authors suggest that many recent tokens built on Ethereum, which used PoW, may help explain this pattern.Source: §8 · pp. 29–30, Figure 16 · p. 32
  • PoW currencies hold about 221 billion USD in market capitalisation and 93% dominance, while PoS and DPoS each have about 6 billion USD and 3% dominance.Excluding Bitcoin and its variants reduces PoW's market-cap share from 93% to 71%, which remains significant relative to PoS and DPoS.Source: §8 · p. 30, Table 29 · p. 30, Figure 17 · p. 32
  • Despite PoW's dominance, the paper reports higher transactions-per-second figures for DPoS currencies than for PoW currencies.Reported figures are Bitcoin at 7 TPS and Ethereum at 15−25 TPS; EOS at 50 reported and 4000 estimated TPS, and Tron at a claimed 2000 TPS.Source: §8 · p. 30
  • The authors discuss first-mover advantage and Bitcoin's market weight as reasons for PoW's popularity, while noting that PoW mining centralisation may also expose it to collusion.They also describe security concerns around smaller validator sets in PoS and DPoS; the potential performance of Ethereum's proposed PoS mechanisms remains untested in real-life settings.Source: §8 · pp. 30–31
  • The decision tree uses incentives, energy consumption, scalability, adversary tolerance, and ASIC resistance to guide algorithm choice across scenarios.The authors present DPoS and BFT derivatives as preferred for incentivised, high-scalability, low-energy scenarios, but describe their security as moderate; prioritizing security points to PoW, with memory-bound PoW trading off scalability and energy.Source: §8 · pp. 31–32
  • The decision tree is an example based on selected criteria, and other criteria can produce a different diagram suited to other scenarios.The authors identify comparison tables across algorithm properties as the template for building such a diagram.Source: §8 · p. 33

9 CONCLUSION

The survey systematically organizes consensus algorithms and blockchain systems, concluding that no algorithm yet avoids significant security or performance disadvantages. It also finds PoW remains dominant by market capitalization and presents a decision tree for choosing algorithms under specified criteria.

9 CONCLUSION

  • The survey introduces a taxonomy of desirable consensus properties and groups algorithms as incentivised or non-incentivised, with incentivised algorithms further divided into three sub-categories.The sub-categories are PoW, PoS, and algorithms beyond PoW and PoS.Source: §9 · p. 33
  • Figure 19 presents a decision tree for choosing appropriate consensus algorithms.The paper describes the tree as a way to test suitability under specified criteria.Source: Figure 19 · p. 34
  • The analysis covers hybrid mechanisms combining PoW or PoS with novel algorithms and non-incentivised mechanisms used in private blockchain systems.The cited examples include Proof of Research, Proof of Burn, Proof of Cooperation, and mechanisms used by Hyperledger systems.Source: §9 · p. 34
  • PoW remains dominant by market capitalization despite its disadvantages, while adoption of PoS and DPoS remains limited.The authors treat market capitalization as an indicator of adoption and note that PoS and DPoS aim to address PoW's limitations.Source: §9 · p. 34
  • The survey excludes non-blockchain distributed ledgers such as IoTA and NANO, whose DAG-based structures and consensus mechanisms are left for future study.The authors state that these systems are outside this article's scope.Source: §9 · p. 35
  • The authors conclude that no consensus algorithm yet avoids significant disadvantages in security or performance.PoW is associated with high energy consumption, low throughput, and poor scalability; PoS is generally fast and scalable but faces attack vectors.Source: §9 · pp. 33–35
Loading 2001.07091v2…