《The Block Research：2024區塊鏈設計報告：協議棱鏡：揭示現代區塊鏈的架構權衡研究報告（英文版）（35頁）.pdf》由會員分享，可在線閱讀，更多相關《The Block Research：2024區塊鏈設計報告：協議棱鏡：揭示現代區塊鏈的架構權衡研究報告（英文版）（35頁）.pdf（35頁珍藏版）》請在三個皮匠報告上搜索。

1、THEBLOCK.COPROTOCOL PRISMUNVEILING THE ARCHITECTURAL TRADEOFFS OF MODERN BLOCKCHAINSBLOCKCHAIN DESIGN REPORTS SERIESCOMMISSIONED BYRESEARCHED BYJANUARY 2024THEBLOCK.CO3JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO2OVERVIEWBlockchains are poised to become a foundational e

2、lement of digital infrastructure.This report focuses on fundamental differences in blockchain design that have been chosen by different protocols to pave the way for mass adoption.Despite all modern blockchains aiming to deliver public rails that are performant and secure,their approaches and timeli

3、nes to achieve those goals vary.To shed light on these differences,we outline the main functions blockchains perform and introduce a framework to categorize them.With case study-type analyses of Ethereum and Cosmos,along with a thorough review of Solanas design approach,we compare their strengths an

4、d weaknesses.Our study highlights the critical differences between horizontal and vertical scaling and the trade-offs between them.This comparison provides a detailed perspective on the technological choices and strategies influencing the evolution of smart contract platforms as the backbone of a gr

5、owing on-chain financial economy.The report is structured as follows:Part 1 defines the core functions of a blockchain system,and then classifies modern blockchains into several encompassing architectures according to the variations across their functions.Part 2 contextualizes the tradeoffs required

6、 for scaling blockchains across each function(such as data storage or execution),highlighting the way in which scalability optimization impacts important properties such as decentralization,security and interoperability.Part 3 takes a deep dive into Solanas design philosophy as an execution-focused,

7、vertically integrated blockchain,as well as the many unique hurdles it has faced in the pursuit of achieving greater scalability.Finally,Part 4 compares the key design choices taken by protocols in the context of their varying scaling approaches,offering insights into their progress to date alongsid

8、e critical commentary from the builders and thinkers shaping the future of blockchain design.SPONSORED BYRESEARCHED BYCONTACTThe Block Pro is The Blocks premium product portfolio designed to help institutions evaluate opportunities in digital assets.Pros research,news,and data products are powered b

9、y teams of subject matter experts deeply entrenched in the digital asset ecosystem who deliver actionable intelligence so businesses can make informed decisions.The Block Research produces research content covering the digital assets,fintech,and financial services industries.Email:researchtheblock.c

10、o Twitter:TheBlockPro_The Solana Foundation is a non-profit organization located in Zug,Switzerland dedicated to the decentralization,advancement,and security of the Solana network.More about Solana Foundation:Website|LinkedIn|TwitterSolana is a global state machine,and the worlds most performant bl

11、ockchain.It gives developers the confidence to build for the long term by delivering predictable scaling without compromising security or composability.Solanas performance is driven by a single global state,which is capable of processing tens of thousands of smart contracts at once,and by Proof of H

12、istory,a distributed clock that unlocks low-latency,sub-second finality across the global state.Established in 2018,Solana Labs is a technology company that builds products,tools,and reference implementations to further expand the Solana ecosystem.More about Solana:Website|LinkedIn|YouTube|Documenta

13、tion|Twitter|Discord|Github|Telegram|Reddit5JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO4ACKNOWLEDGMENTSTABLE OF CONTENTSWe would like to thank the Solana Foundation for commissioning this research report.We would also like to thank their team for providing feedback and

14、 input for this report,in particular Austin Federa.We would also like to thank everyone at The Block who assisted with this report-design team:Zoe Ellyse Del Rosario;research team:Marcel Bluhm,George Calle,Greg Lim,Arnold Toh.We are also grateful to those that shared their valuable perspectives thro

15、ugh interviews for this report:Sunny Aggarwal(Osmosis Labs)Lucas Bruder(Jito Labs)Noam Cohen(Binary Builders)Justin Drake(Ethereum Foundation)Austin Federa(Solana Foundation)Haseeb Qureshi(Dragonfly Capital)Kyle Samani(Multicoin Capital)Jeff Washington(Solana Labs)Anatoly Yakovenko(Solana Labs)The a

16、uthors of this report may hold tokens mentioned in this report.Please refer to The Blocks financial disclosures page for author token holdings.AUTHORKevin Peng,PhDResearch AnalystPengCapitalLinkedInINTRODUCTION6PART 1|CLASSIFYING MODERN BLOCKCHAINS81.1 MONOLITHIC BLOCKCHAINS101.1.1 RISE OF EVM111.1.

17、2 PUSHING THE LIMITS OF EVM-COMPATIBLE L1S141.2 MONOLITHIC BLOCKCHAINS WITH PARALLEL PROCESSING151.3 MULTICHAIN PROTOCOLS181.4 MODULAR ARCHITECURES201.5 SHARDED BLOCKCHAINS AND THE FUTURE OF SCALING24PART 2|SCALABILITY TRADEOFFS ACROSS BLOCKCHAIN FUNCTIONS282.1 DATA STORAGE302.2 CONSENSUS AND SETTLE

18、MENT312.3 EXECUTION34PART 3|SOLANA:THE CASE FOR MONOLITHIC EXECUTION383.1 OPTIMIZING EXECUTION FROM THE GROUND UP393.2 IMPROVING VALIDATOR PERFORMANCE:EXPERIMENTATION IN PRODUCTION42PART 4|MODULAR AND MULTICHAIN VS.MONOLITHIC SCALING:CURRENT STATE AND FUTURE ROADMAPS504.1 THROUGHPUT AND EXECUTION524

19、.2 HARDWARE REQUIREMENTS AND DECENTRALIZATION564.3 ONWARD AND UPWARD:THE STATE OF BLOCKCHAIN SCALING AND BEYOND60CONCLUSION667JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO6Modern blockchains have evolved dramatically from their humble origins as simple peer-to-peer(P2P)n

20、etworks for the secure transfer of money.Smart contracts,first introduced on Ethereum,have played an especially key role in enabling blockchains to expand their capabilities far beyond processing P2P transactions.With the rapid growth of decentralized finance(DeFi)protocols built upon smart contract

21、s in recent years,including decentralized exchanges(DEXs),lending protocols,synthetic asset protocols,non-fungible token(NFT)marketplaces,and more,crypto users now have unprecedented access to a stunning array of mostly permissionless financial instruments.Today,tens of billions of dollars worth of

22、crypto assets are held in DeFi protocols alone.The smart contract-based blockchains that host these protocols are home to an even greater amount of capital overall,by an order of magnitude.As of November 2023,about$125 billion worth of fiat-backed stablecoins such as USDC and USDT reside on smart co

23、ntract platforms.In October 2023,$390 billion in stablecoin volume was settled on the Ethereum blockchain alone,underscoring the immense trust that is now placed in smart contracts and their underlying blockchains.This expanded optionality with respect to on-chain financial activity would not have b

24、een possible without a corresponding redesign of popular blockchain architecture in the early years following Bitcoins rise to prominence.For instance,the introduction of the Ethereum Virtual Machine(EVM)and its Turing complete state machine laid the foundation for the smart contracts that underpin

25、most crypto applications today.In the current blockchain landscape,core development teams have continued to design and build new functionalities for blockchains,primarily with the goal of enhancing scalability.After the sudden surge in both the volume and complexity of DeFi transactions on-chain beg

26、inning in late 2020,it has become especially clear to users and developers throughout the crypto ecosystem that blockchains still require significant improvements before they will be able to effectively meet the demands of a global digital economy.However,one of the major challenges with evaluating

27、the state of blockchain development today is the sheer variety of design goals and approaches that exist among various chains.Some chains exist solely to execute transactions while others are intended to provide security guarantees for other chains.In many cases,different development teams will use

28、contrasting language to describe similar concepts,depending on their particular implementation or focus.In this report,we aim to provide an in-depth understanding of the modern blockchain landscape in the context of the varying design philosophies that exist today.First,we provide an objective frame

29、work for classifying blockchains according to their key functions and architecture.Then,we explore some of the main components that underlie network functionality,as well as the various trade offs made within different initiatives to optimize these components.Finally,we conduct a deep dive into the

30、Solana networks design and scaling approach,along with an assessment of its progress in comparison to other popular approaches today.Lets begin with a look at different blockchain functions and how we can use them to categorize nearly all blockchains today.INTRODUCTION9JANUARY 2024BLOCKCHAIN DESIGN

31、SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO8PART 1CLASSIFYING MODERN BLOCKCHAINSAt their core,fully-fledged smart contract platforms must all perform the same key responsibilities:transaction execution,settlement,consensus,and data availability(DA).These are often referred to as different“layers”in

32、 a hypothetical blockchain with fully modular components.Briefly,the execution layer is responsible for executing state transitions within a particular execution environment or virtual machine.In other words,execution defines nearly all aspects of user interactions with a specific blockchain on a fu

33、ndamental level,from the way transactions are sent between addresses to the way contracts are written to perform certain actions.The settlement layer is where transaction correctness and finality are established;transaction correctness is established through methods such as validity proof verificati

34、on,zero-knowledge proofs,and cryptographic signatures,or via dispute resolution(a.k.a.social consensus).Finality is achieved through consensus mechanisms that include both social consensus and algorithmic consensus,such as Proof-of-Work(PoW),and Proof-of-Stake(PoS).Meanwhile,the consensus layer is r

35、esponsible for-as the name suggests-achieving consensus between various nodes/validators on the order of transactions.The data availability layer is responsible for attesting to and providing transaction data availability-either to other layers or independent users querying the data.In many blockcha

36、ins today,such as on Ethereum or Avalanche,the settlement layer is coupled with the consensus layer or the data availability layer,rather than existing as a standalone function.11JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO10These key blockchain functions execution,sett

37、lement,consensus,and data availability form the basis of our framework for classifying modern blockchains.This framework is sometimes referred to as a“modular”framework,which,along with the inclusion of the data availability layer,was largely popularized as a term by the Celestia team in recent year

38、s.It is worth noting the potential for some implicit bias with the use of these terms,given the fact that Celestia intends to be a leading data availability layer for blockchains with a rollup-centric roadmap.For instance,the terms“monolithic”and“modular,”taken together as terms used to describe blo

39、ckchains,can easily be misconstrued to imply that monolithic blockchains are non-modular and single-function,while modular chains are flexible and multi-functional.The truth is often far more complex.In fact,some would argue that the popular definition of a monolithic chain,i.e.one that performs all

40、 blockchain functions,might be more suitable for so-called modular layers like Celestias data availability network,which only publishes transactions but does not execute them.Nonetheless,we still find it useful to group and classify blockchains by the way they perform(or dont perform)key functions,a

41、nd largely stick with the common definition for monolithic chains in this report.We also retain the data availability layer in our framework,given the growing acceptance of DA as a function for scaling rollups.In terms of our classification framework,we prefer to describe it as a blockchain function

42、 framework,as opposed to a modular framework,to more accurately reflect our focus on clearly defining blockchains by their responsibilities/functions,rather than suggesting that these responsibilities must be optimized individually.1.1 MONOLITHIC BLOCKCHAINSMost blockchains throughout history have b

43、een monolithic chains,performing all key responsibilities outlined above within a single architecture.The first blockchain,the Bitcoin blockchain,along with other early blockchains such as Litecoin and Dogecoin,are examples of monolithic blockchains.Notably,all three of these blockchains do not have

44、 smart contract functionality,and are thus limited to crypto transfers between peers on the network.Smart contracts,first introduced on Ethereum,enable a far wider range of transaction types and user behavior,but require robust Turing complete execution environments,which dictate execution logic.Sma

45、rt contract platforms like Ethereum,Avalanche,and Solana are commonly referred to as monolithic chains as well,despite the fact that they must execute more complex transaction logic and bear more computational load.In this report,we refer to monolithic chains solely in the context of smart contract

46、platforms,which now serve as the backbone for the bulk of decentralized financial activity on blockchains today.Over time,smart contract functionality has given rise to increasingly sophisticated DeFi protocol design,providing blockchain users with new degrees of financial flexibility that have,in t

47、urn,created new channels of demand for block space.It is the confluence of these forces that has pushed early smart contract platforms like Ethereum to the limits of their natural capacity.This brings us to the center of where a significant fraction of core blockchain innovation lies today:scaling.I

48、n general,monolithic blockchains also referred to as Layer 1(L1)networks-perform the full range of key blockchain functions within their own architecture,but that is roughly where their similarities end.In the pursuit of greater scalability,some L1s have begun to adopt a modular approach to developm

49、ent,embracing the idea of separating key functions into separate chains to be optimized on an individual level.Layer 2(L2)networks,commonly referred to as rollups,embody this philosophy by focusing solely on execution to segregate computational load,leaving the remaining functions to an underlying L

50、1,most commonly Ethereum.Other L1s,such as Solana,have taken a different approach,aiming instead to maximize the performance of their existing functions within a single,monolithic architecture.As it turns out,when it comes to execution,regardless of architecture,it is the underlying execution enviro

51、nments that largely give way to variations in overall performance.1.1.1 RISE OF THE EVMFirst and foremost,Ethereum aspires to be the blockchain with highest(economic)security.And then,number two,we want to have enough scalability to provide a lot of utility to the world.Justin Drake(Ethereum Foundat

52、ion)The EVM is by far the most dominant execution environment among both L1 and L2 smart contract platforms today;utilized by 9 of the 10 largest blockchains by total value locked(TVL),as of December 2023.The explosive growth of EVM-compatible chains in recent years largely stems from Ethereums pivo

53、tal role in fostering many of the core DeFi primitives still used today.In 2020,during what became known as“DeFi Summer,”the combined market cap of DeFi protocols grew to a then-record high of$20.8 billion,capturing 5.3%of cryptos total market cap at the time.By the end of 2020,nearly all the TVL an

54、d activity in DeFi remained on Ethereum,with a 96%market share firmly establishing the L1 network as the leader in the new DeFi space.13JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO12Over the ensuing years,Ethereum would go on to lose a significant portion of its DeFi ma

55、rket share to alternative L1s,even as DeFis total market cap reached staggering new heights exceeding$170 billion at its peak.Buoyed by rampant speculation and record inflows to crypto markets,the events of 2020-2022 were ultimately highly instructive for identifying the key challenges in modern blo

56、ckchain scaling,as well as the strengths and limitations of the EVM when bounded by a monolithic architecture.One of the strongest drivers of the adoption of EVM-compatible L1s was the Ethereum networks pronounced struggles with surging user demand throughout 2021.Average transaction fees on Ethereu

57、m soared to record highs in the first half of 2021,coinciding with the rapid growth and initial adoption of Binance Smart Chain(now known as BNB Chain).15JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO14BNB Chains out-of-the-box compatibility with the EVM meant that the gr

58、owing community of Ethereum users and developers could seamlessly transition to the new L1 without significant changes to their typical workflow,such as accessing protocols via Metamask.With BNB Chain touting fast confirmation times and low transaction fees,the additional convenience of its EVM supp

59、ort proved to have a profound effect.By May 2021,the BNB Chain ecosystem comprised over 20%of total DeFi TVL.That same year,with a little help from liquidity mining incentives,several other EVM-compatible L1s would go on to see unprecedented growth while offering a similar value proposition as BNB C

60、hain,with Avalanche,Fantom,and Polygon capturing 5.8%,2.4%,and 3.0%,respectively,of DeFi TVL by the end of 2021.1.1.2 PUSHING THE LIMITS OF EVM-COMPATIBLE L1SThere are several reasons why,compared to Ethereum,transactions tend to be cheaper on L1s such as BNB Chain,Avalanche,or Fantom,all of which a

61、re EVM compatible and have a similar monolithic architecture.One of the main reasons is that demand for block space on these L1s is simply not as high relative to Ethereum.Even today,Ethereum remains dominant among smart contract platforms in terms of TVL,DEX volume,and NFT market share.As such,it b

62、ecomes difficult to use nominal transaction fee amounts as reliable indicators of scalability in direct comparisons between L1s.In the past few years,isolated spikes in demand for specific L1s have yielded spurts of useful insights into how EVM-based L1s realistically perform under duress.For exampl

63、e,in March 2022,the Fantom network saw its average gas prices temporarily skyrocket to record highs after one of its core devs abruptly announced his departure from the industry.Other notable spikes in gas prices include May 2022 during the collapse of Terra Luna,November 2022 during the collapse of

64、 FTX,and most recently,July 2023 during the shutdown of the Multichain bridge.All the while,average daily transactions have been on a steady decline since early 2022.These incidents highlight the fact that despite offering lower fees compared to Ethereum on average,Fantoms design does not make it im

65、mune to the negative effects of sudden increases in block space demand.Similar conclusions can be drawn with other monolithic chains that utilize the EVM,most of which exhibit the same gas profile as Fantom.Sensitivity of gas prices to block space demand is one indicator of a networks overall scalab

66、ility,though organic demand is generally tricky to measure or reproduce in a purely mathematical sense.One study conducted by Gengmo Qi and the Dragonfly Research team sought to quantify the throughput of various L1s in terms of DEX trades,factoring in the block gas limit,block time,and gas expendit

67、ure of a typical token swap on each chain in order to determine average swaps per second.Notably,the team found that EVM-compatible L1s were largely bound by the same scaling constraints due to using the same gas model and state transition function.For instance,BNB Chain is able to process far more

68、trades per second relative to its peers,but it also makes significant tradeoffs in order to do so-as of this writing,the network still has only 29 active validators.The paper also suggests that users appear more interested in ecosystem strength,user experience,and low fees than in performance on non

69、-Ethereum L1s.Performance isnt a competitive factor(yet)as these blockchains arent consistently capacity constrained,barring occasional usage spikes.Further discussion on scaling tradeoffs in general can be found in Part 3 of this report.1.2 MONOLITHIC BLOCKCHAINS WITH PARALLEL PROCESSINGA common ch

70、aracteristic of most smart contract platforms today is that they utilize a single-threaded runtime,as in the EVM.In other words,transactions are processed sequentially,ensuring that conflicting transactions are not executed during state transitions.However,this behavior can also impose a limit on tr

71、ansaction throughput,especially when considering that transactions are often submitted simultaneously and much more frequently than they are being finalized on chain.One of the main thrusts to directly address these throughput bottlenecks is the development of blockchains that employ parallel proces

72、sing of transactions.If early non-smart contract blockchains such as Bitcoin are considered as“1st-gen”chains,and smart contract platforms such as Ethereum are considered“2nd-gen”chains,then blockchains with parallel processing capabilities can be reasonably categorized as“next-gen”for the theoretic

73、al benefits they provide.“Because of Solanas speed and low fees,you have more flexibility in the type of DApps that you can create.”Lucas Bruder(Jito Labs)The most prominent example of this design strategy in the market today is the Solana network,which grew at a meteoric pace alongside its EVM-base

74、d L1 competitors in 2021 before cooling down in the aftermath of the FTX collapse in 2022.Instead of the EVM,Solana utilizes a custom execution environment known as the Solana Virtual Machine(SVM).17JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO16The key component of the

75、SVM that enables parallel processing is the Sealevel runtime,which requires all transactions to explicitly define which accounts will be read from or written to.In practice,this adds an additional level of complexity for Solana developers,but it also enables an important step change in terms of tran

76、saction processing efficiency.Since transactions that do not introduce conflicting state changes are clearly marked prior to execution,they can be effectively grouped together and processed simultaneously,i.e.in parallel.For example,if account A sends 1 SOL to account B,this should have no impact on

77、 whether account C can send 1 SOL to account D;thus,processing these two transactions sequentially-as in the EVM-can be considered a source of inefficiency.In order to fully leverage the efficiency gains imparted by the SVM,Solana relies on several additional features that are specific to its monoli

78、thic architecture,such as Pipelining transaction processing unit(TPU)and its Proof of History(PoH)synchronization system.At a high level,this approach can be summarized as an extreme version of vertical scaling with its own unique tradeoffs,which are discussed in further detail in Part 3 of this rep

79、ort.In recent years,other core teams have begun to deploy blockchains that feature a similar focus on parallel processing as a means of enhancing the scalability of monolithic systems.The most notable examples of this are Aptos and Sui,two L1s that were spun off from Metas original Diem blockchain p

80、roject.Aptos and Sui both use parallel processing mechanisms that bear similarities to Solanas.In Aptoss Block-STM engine,state-independent transactions are processed with a preset order as determined by its consensus engine;Suis Narwhal/Tusk consensus engine performs a similar data ordering task th

81、at ultimately reduces computational load prior to transaction execution.Perhaps the most notable trait shared between Aptos and Sui is their use of the Move virtual machine(Move VM),which requires contracts to be written in the Move programming language.Similar to the SVMs support for Rust and C/C+o

82、ver Solidity,the Move VMs unique programming requirements pose an additional barrier for existing EVM developers looking to transition to Aptos or Sui.This circumstance is largely derived from the fact that novel execution environments tend to require novel instructions as well.More importantly,it i

83、s clear from these examples that execution environments play a major role in dictating user experience,and are currently one of the key variables for enhancing transaction throughput as well.19JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO181.3 MULTICHAIN PROTOCOLS“The st

84、rength of Cosmos is that you can build your own application.If you want customizability on top of sovereignty,speed and decentralization,Cosmos is the way to go.”Noam Cohen(Binary Builders)In the previous section,we reviewed the current state of efforts to maximize throughput on individual smart con

85、tract platforms,where we found that execution environments are a central bottleneck for scaling.Whereas parallel processing is intended to improve the performance of monolithic blockchains under heavy demand,some protocols have taken an alternate approach to combatting congestion,namely,by separatin

86、g user applications into individual blockchains.This strategy was largely popularized through the Cosmos teams original vision of an“internet of blockchains,”which is essentially a collection of application-specific chains(app-chains)that are connected by a shared communication protocol.Aside from t

87、his dedicated communication standard,namely the Inter-Blockchain Communication(IBC)protocol,blockchains in the Cosmos ecosystems do not differ meaningfully from other monolithic chains in terms of their performed functions.With few exceptions,other“multichain”protocols are essentially the same;they

88、perform their own execution,settlement,data availability sampling,and consensus.As such,they are most accurately described as“multichain monolithic”networks,and we focus primarily on the slight variations in their architecture in this section.The general idea behind Cosmos is that application develo

89、pers should be able to easily create their own blockchains that are tailored to their specific use cases.Cosmos chains are built via the Cosmos SDK and utilize the Tendermint BFT consensus engine,but they remain entirely sovereign networks.Sometimes referred to as“hubs,”these networks typically feat

90、ure their own validator sets,along with independently variable governance parameters and security guarantees.Meanwhile,the IBC protocol acts as a standardized means through which Cosmos networks can securely pass messages between one another,further enabling cross-chain asset transfers and general U

91、X integration between applications.It is important to note that transactions are settled on their respective chains,meaning that the confirmation of blocks on one chain has no direct impact or relation to other chains.One of the benefits of this architecture is that computational load is split among

92、st various chains depending on user demand at specific times.In addition,governance or consensus failures on one app-chain would be confined to that particular chain,as opposed to the broad impacts of a similar failure on a monolithic chain featuring many applications and protocols.However,multichai

93、n protocols also face issues of fragmented resources;each chain is tasked with bootstrapping enough human and economic capital to establish a secure validator set,and liquidity is generally fragmented across multiple chains.Recent developments in the Cosmos ecosystem have aimed to address these chal

94、lenges with features like Replicated Security,which would allow new“consumer chains”to essentially rent security from the Cosmos Hub chain by using its validator set.Under this design,Cosmos Hub validators perform all validation for consumer chains,collecting 100%of fees generated from the consumer

95、chain.Of course,this also requires general social consensus on the Cosmos Hub serving as a settlement layer.It is important to note that settlement and consensus for consumer chains under the Replicated Security model remains separate from the Cosmos Hub;consumer chains receive periodic IBC packets

96、to update their validator sets to match the Cosmos Hub,but activity on consumer chains still has no relation to activity on the Cosmos Hub,and vice versa.As of now,Hub validators are compelled to perform validation for consumer chains that have been onboarded to the Replicated Security model through

97、 governance,but there is a technical limit for how many chains Cosmos Hub validators can reasonably support.In the future,these validators would be able to“opt-in”to consumer chain validation,collecting fees only from those they choose to support.Avalanche has employed a similar horizontal scaling a

98、pproach with subnets,which are essentially independent chains that are constrained by the requirement of utilizing a subset of Avalanche C-Chain validators.This adds an additional layer of security to subnets,as subnet validators must already be honest participants on the C-Chain,but it is still lim

99、ited by the size and economic security of the particular validator set.This is similar to the opt-in version of the Cosmos Hubs Replicated Security model,whereby validators 21JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO20for the more popular C-Chain can perform validati

100、on for subnets,but consensus achieved on the C-Chain does not directly translate to consensus on subnets.Polkadot is another multichain protocol that effectively delegates applications to individual blockchains,with the main differentiator being that all blockchains in its ecosystem,known as paracha

101、ins,are secured by a single governing chain called the Polkadot Relay Chain.The overall Polkadot network is heavily dependent on the functioning of its Cross-Consensus Message(XCM)protocol,which facilitates both Relay Chain-parachain and parachain-parachain communication.As is the case with all mult

102、ichain protocols,the cross-communication standard is ultimately an enabler but also the main limiting factor when it comes to overall throughput,security,and liquidity.Among multichain protocols,the Polkadot model is perhaps the closest to a true“shared security”model,where parachain transactions mu

103、st also be confirmed on the primary Relay Chain.However,we have also discussed at length in previous reports how this arrangement has effectively bound the pace of development on parachains to the current state of development on the Relay Chain.1.4 MODULAR ARCHITECTURESIn some ways,multichain protoc

104、ols can be interpreted as an extension of modular architecture,wherein the design usually offers a choice between either siloed,application-specific security,or delegated security to a single consensus layer.The rollup-centric roadmap laid out by the Ethereum foundation has drawn considerable scalin

105、g efforts towards the idea of Ethereum being the“global settlement layer”,leading to the proliferation of various L2 solutions.L2s,which can be further categorized by their validation mechanisms and data availability modes,are a fundamental extension of the basic premise that effective blockchain sc

106、aling can best be achieved by optimizing individual blockchain functions.As mentioned above,these are generally defined as execution,settlement,consensus,and data availability layers in a theoretical modular blockchain.Since transaction processing and execution are computationally intensive tasks,ro

107、llups today generally exist for the purpose of separating execution from the settlement and consensus layers.In other words,rollups represent the execution layer(L2)for an underlying L1 such as Ethereum.The general mechanism for rollups is as follows:transactions are executed on a rollup chain,which

108、 are subsequently batched by a sequencer,producing a snapshot of the resultant VM state,which is then sent to the settlement layer for verification and consensus layer for finalization.Part 4:Conclusions&OutlookL2s that leverage fraud proofs differ greatly from those that leverage validity proofs in

109、 terms of their security assumptions.Examples of L2s using fraud proofs include Arbitrum,Optimism and Base and examples of L2s leveraging validity proofs include Starknet and Loopring.The former assumes that submitted batches are correct and waits for an arbitrary dispute period before finalization

110、while the latter necessitates a validity proof that guarantees the correct execution of the proposed batchs transactions.Typically,validity proofs are to be submitted with every L2s state update,thereby consuming more computational resources on the underlying layer,Ethereum,than L2s using fraud proo

111、fs,since fraud proofs are only required in the event of a dispute.That said,validity proofs allow for state finality to be achieved much faster than fraud proofs.There is also a growing opportunity in providing alternative data availability(DA)solutions for L2s,as rollups typically rely on the under

112、lying L1,Ethereum,for data availability.This is further evidenced by the growth of Validiums like ImmutableX and Optimiums such as Mantle,which leverage off-chain data availability solutions.It should be noted that DA differs from data storage,as DA only requires that data can be accessed when queri

113、ed while data storage refers to having the complete archive of past transaction data.Simply put,it is possible for one to make a DA layer much lighter than a data storage layer,in exchange for a marginal decrement in the guarantee of the data being accessible.23JANUARY 2024BLOCKCHAIN DESIGN SERIES:P

114、ROTOCOL PRISMSPONSORED BYTHEBLOCK.CO22There are currently two types of DA layers:centralized DA committees,which are typically used by StarkEx protocols,and independent DA-optimized layers,such as Celestia and Avail(prev.Polygon Avail).The former comprises of multiple entities providing attestations

115、 that they possess a copy of proposed batches transaction data while the latter submits random data samples on-chain to attest that the batched transaction data is available with high probability.Both approaches aim to alleviate the costs associated with storing all transaction data on-chain and it

116、shows some extent of efficacy when comparing gas costs associated with rollups using on-chain DA against off-chain DA.The chart below compares the gas consumption of Ethereum Layer 2s leveraging validity proofs with on-chain DA(ZK rollups)against off-chain DA(Validiums).However,the progress towards

117、increased modularization and abstracting DA is not without costs.The debate between modularization and integration is not entirely novel;it had been present when operating systems tech stacks were being developed.Fundamentally,modularization does not enhance the scalability of any blockchain.Instead

118、,it can be argued that it shifts the burden of computation to different entities and when demand for computation surges,the problem resurfaces.For example,both Arbitrum and Optimism saw sustained periods of high gas fees when Ethereum gas fees were consistently high.In other words,modularization can

119、 optimize how resources are utilized and potentially improve performance under certain conditions,it doesnt eliminate the fundamental scalability challenges facing blockchain networks.Its a step toward managing scalability issues,not a complete resolution.Understanding data availability is fundament

120、al to designing a blockchain that scales well because it is the main bottleneck for most,if not all,public blockchains today.Specifically,blockchain nodes must be certain that the data in a block is accessible if it is to accept the block as valid.As such,various approaches have been experimented wi

121、th,from on-chain data availability,which are used by most Layer 1s,to off-chain data availability,employed by Validiums and Optimiums(previously known as Plasma),to data availability optimized layers,such as Celestia.As the need for increased throughput rises,it will be inevitable that we see more p

122、ublic blockchains today adopting some form of data availability optimization in order to keep up with the demand for block space.A handful of Layer 1s aim to tackle this problem from the ground up,by integrating localized fee markets,such that users interacting with high-demand smart contracts pay m

123、ore gas than users making a simple transfer.This provides a targeted incentive to lower demand for the frequent users.It remains to be seen how such novel incentive/pricing mechanisms play out in the market when demand is back and blockchain resources become scarce again.25JANUARY 2024BLOCKCHAIN DES

124、IGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO24To sum up,typical L1s have assumed the four responsibilities of execution,settlement,consensus and data availability.L2s have attempted to outsource both execution and data availability.While execution has proven to increase the throughput of the und

125、erlying L1 to some extent,evidenced by the growth in adoption of optimistic and ZK rollups,Validiums like ImmutableX and ApeX have shown more potential to scale blockchains.That said,Validiums make some trade-off in security by hosting data off-chain,which is why we see the rise of data availability

126、 optimized layers like Celestia.Based on the diagram above,we are about to reach the very edge of how modular a blockchain can be,which means the main focus will return to the fundamental design of a blockchain.1.5 SHARDED BLOCKCHAINS AND THE FUTURE OF SCALINGSharding is a technique used to improve

127、the scalability and performance of blockchain networks by dividing the networks workload into smaller,manageable pieces,each called a shard.It can be seen as running a blockchain as parallel chains that merge every few blocks to establish a consensus on the global state,before dividing into multiple

128、 chains again for processing transactions.Sharding is an extremely complex approach to scaling that seeks to increase a blockchains throughput without sacrificing security.The current state of development,however,reveals the difficulty and nuance involved in either building net new sharded blockchai

129、ns or implementing sharding within existing blockchain architectures.The complexity and scalability benefits vary depending on what layer of the blockchain stack is being sharded.Blockchains can either have a monolithic execution chain with shards used only to store data(data shards),or have nearly

130、independent shards(execution shards),an approach that is marginally different from running separate Proof-of-Stake chains.It should be noted that data shards are different from data availability optimized layers,as data shards store transaction data in its entirety.As such,data shard nodes are expec

131、ted to hold said data and ensure that the data remains accessible for all network participants.While data shards are less challenging to implement,they do little to improve the chains scalability unless accompanied by rollups,since L1 transaction data can be stored on the main chain.Thus,data shards

132、 are primarily meant as an alternative for rollups to post calldata,which then introduces the issue of the liveness of shards whether shard nodes are able to provide transaction data upon request as readily as full nodes.This is crucial because rollups require transaction data to be accessible when

133、verifying a validity proof or a fraud proof.Execution shards,which can execute transactions on their own,with their own set of validators,are much more challenging to implement.This is expected considering how every shards state must periodically align with a global consensus each time the shards me

134、rge.It is similar to a situation with multiple chains,each with its own consensus,arriving at a global consensus every few blocks.Additionally,execution shards require significant economic stake to secure,as they become capable of proposing fraudulent state transitions,giving malicious actors an inc

135、entive to compromise individual shards.However,neither splitting stakers across shards nor sharing a common stake across all shards would be optimal.Splitting stakers across shards would mean fragmenting the economic stake securing the whole chain,thereby reducing the economic security of a single s

136、hard,while sharing a common stake would mean that stakers are expected to validate the actions of all other stakers,essentially giving stakers the same computational work as a single unsharded chain.Still,there have been numerous attempts at building shard chains,such as Near,Harmony and Elrond.Thes

137、e shard protocols all share one commonality:smart contracts are hosted on a single shard.This is because cross-shard smart contract interactions are significantly more complex,with more potential vulnerabilities.To understand why,we can look at cross-chain smart contract interactions.Typical blockch

138、ain interactions are atomic;either the whole transaction happens,or not at all,while cross-chain interactions are not atomic.For most smart contract interactions,transactions need to be atomic.For example,a token swap on a decentralized exchange must either swap token A for token B,or make no swap.A

139、 break in atomicity,such as sending token A in the first transaction and receiving token B in the next transaction,increases the susceptibility for malicious attackers to revert the second transaction.As a result,when performing a cross-chain swap,the user trusts the facilitator to behave honestly.I

140、n the case of cross-chain decentralized exchanges,there is typically a centralized actor or a centralized group of actors.However,a permissionless shard chain protocol like Near or Elrond would not be able to ensure that all nodes would honestly facilitate a cross-shard interaction,and the incentive

141、 design for such a responsibility is still not well-established,especially since a large cross-shard swap could be highly profitable for a malicious node to intercept.Thus,the design for execution shards is still incomplete and may not be ready in the foreseeable future.27JANUARY 2024BLOCKCHAIN DESI

142、GN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO26In the case of Ethereums Proto-Danksharding,data blobs are attached to Ethereum blocks to serve as cheaper alternative for on-chain data storage for L2s.These data blobs are also automatically discarded after some time,ideally within 3 months.This wou

143、ld be made even cheaper when Ethereum launches its shard chains,allowing for up to 64 different data blobs to be attached.29JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO28PART 2SCALABILITY TRADEOFFS ACROSS BLOCKCHAIN FUNCTIONSIn the previous part,we looked at the general

144、 range of architectures for modern blockchains,tracing their evolution from simple p2p networks used primarily for the transfer of a native asset,to the complex,dynamic smart contract platforms that underlie on-chain financial ecosystems today.The rising computational demands of this emerging digita

145、l economy have driven most blockchain developmental efforts toward a single common goal:scaling.This landscape has largely informed our framework described above,which classifies blockchains according to their design with respect to key functions:execution,settlement,consensus,and data availability.

146、In recent years,growing efforts have been made toward individually optimizing these different functions of a blockchain,particularly in the Ethereum community,in the hopes of enhancing blockchain scalability overall-a so-called modular approach.Still,optimizing even a single function is a non-trivia

147、l task.In this section,we break down the main network components that underlie these functions,as well as the various tradeoffs made in the name of enhancing a particular functionality.As part of this discussion,it is useful to consider the impact of various blockchain functions through the lens of

148、a popular mental model:the scalability trilemma.The scalability trilemma describes the interdependent relationship between scalability,decentralization,and security in blockchain architecture.This dynamic serves as a general design constraint for scaling efforts,and essentially states that maximizin

149、g two of the three factors inevitably leads to decreased competency in the third.Each of these three properties is critical for blockchains to function at any point in time.Scalability generally refers to the ability of a network to process a high volume of transactions per second beyond the capabil

150、ities of a single consumer node.Decentralization is analogous to censorship resistance,which means that a properly decentralized network places little to no trust in a small group of nodes that could potentially fail or become compromised,impacting the entire network.Finally,security refers to a net

151、works ability to resist an economic attack either via collusion between participating nodes or a direct takeover of a controlling stake in the network.Early efforts to scale blockchains have illustrated the difficulty in directly increasing throughput without having deleterious effects on decentrali

152、zation or security.For instance,simply increasing block size can produce short-term benefits in terms of throughput,but ultimately imposes higher transaction costs on nodes,which has a net negative effect on decentralization by limiting the number of potential participating nodes.Similar challenges

153、quickly arise when designing blockchains that maximize one or two properties without also solving for resultant weaknesses in the third.A highly scalable,decentralized blockchain faces 31JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO30heightened security risks due to rela

154、tively lower costs of attacking the network.By the same token,highly decentralized,secure blockchains are often harder to scale,given that throughput generally becomes throttled by higher latency between a larger number of nodes.Ultimately,the scalability trilemma should not be thought of as strict

155、limiting factor for blockchain development,but more as a general mental framework to understand the tradeoffs that result from enhancing specific functions.Over time,this framework has become progressively less imposing on blockchain design as teams have continued to produce breakthroughs that are a

156、ble to effectively mitigate the issues in past designs.Even so,the scalability trilemma still remains generally relevant as a model when considering blockchain design approaches as a whole,especially as teams continue to experiment with new ways to push their networks beyond their current scaling li

157、mits.Below,we review some of the underlying components that comprise the key blockchain functions in the context of the scalability trilemma,focusing on how these underlying components either directly or indirectly influence the delicate balance between scalability,decentralization,and security.2.1

158、DATA STORAGEData storage is a critical part of blockchain functionality,and can be roughly divided into global state and historical data.Global state can be thought of as a live snapshot of all the data that can be accessed and modified for a blockchain at a given time.This data is typically stored

159、in Merkle trie structures,which allows it to be quickly retrieved by validators for verification and execution of the next state change.On the other hand,historical data refers to the full range of data generated throughout the course of a blockchains lifespan,including all confirmed blocks,all the

160、transactions within those blocks,transaction signatures,etc.Generally,historical data is needed to sync new validators to a blockchain,but does not need to be frequently accessed by active validators who rely on global state data to execute state transitions.However,over time,the constant buildup of

161、 raw data can lead to growing storage demands,progressively increasing the computational and economic load on validators.This is commonly referred to as state bloat,which can have a negative impact on transaction execution and throughput if validators begin to take longer to confirm transactions.At

162、the same time,increasing hardware requirements for validators to combat state bloat can become a centralizing force,limiting the number of participants capable of running a validator.One of the primary ways that historical data can be stored for EVM-based blockchains,which can have significant amoun

163、ts of state bloat,is to use off-chain storage solutions.These solutions are typically blockchains themselves,offering a basic level of decentralization,but they also face the same issues that are relevant to smart contract platforms.For instance,they require some sort of incentive structure to ensur

164、e that nodes continue to write,store,and provide access to data,and they also require the ability to scale themselves.Some solutions,like Arweave,reward validators for adding new blocks,similar to Ethereum.Others,like Filecoin or Storj,rely on a contract-based mechanism for storing specific data ove

165、r a given period of time,which enhances their scalability by reducing the complexity and persistence requirements of data storage.Still,it is important to note that most blockchains,including Ethereum,have not yet enshrined data storage solutions as a part of their protocol,and continue to rely on n

166、odes being able to store progressively larger amounts of historical data over time.Data storage is also important in the context of enabling rollups,which require the submission of either fraud proofs or validity proofs to the underlying L1 that can be accessed at any time.Note,however,that data sto

167、rage typical refers to blockchain historical data in the context of state bloat,while data availability more specifically refers to the ability to access required data as needed,without which rollups could not be considered secure.As discussed in Part 1.3,rollups typically utilize one of two types o

168、f data availability layers,which guarantee data access on demand rather than full archival of historical data.These include centralized data availability committees and independent data availability layers,each with their own economic and security tradeoffs.2.2 CONSENSUS AND SETTLEMENTThere are seve

169、ral key factors that ultimately contribute to a networks ability to consistently achieve consensus and settle transactions.One of the most important is governance,which is often overlooked but can have sweeping implications for a particular networks security and future development.At a high level,go

170、vernance dictates the overall process for network upgrades and serves as a coordination mechanism between users,developers,validators,and stakeholders.In the event of potential network failures or malfunctions,governance can also be the ultimate adjudicator for consensus disputes.Modern smart contra

171、ct platforms feature a wide range of governance structures as part of their overall design.Ethereum,for example,employs an off-chain governance structure that is mostly conducted 33JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO32through social consensus involving various

172、stakeholders.Despite the fact that the network is the largest and likely most decentralized among smart contract platforms today,its governance has been relatively effective at enacting necessary changes to the network over time.Ethereums successful transition from Proof of Work(PoW)to Proof of Stak

173、e(PoS)consensus in September 2022,albeit delayed,is a testament to this functionality,notably revamping its network architecture despite initial opposition from now-defunct Ethereum miners.Decentralization of governance is important for maintaining a networks overall resistance to censorship,but it

174、can sometimes be a barrier to timely protocol implementation as well.It is worth noting that despite Ethereums eventual transition to PoS in 2022,many of its L1 competitors had already adopted the mechanism since their inception several years before.Some networks have established a more formalized g

175、overnance structure by conducting the process on-chain.For instance,Cosmos chains delegate voting powers to validators,who have the ability to set custom governance parameters(voting period,quorum,veto threshold,etc.)depending on their individual needs.In theory,this design has the effect of alignin

176、g interests between governance and stakeholders,as validators are often among the largest stakeholders in the network as well.Thus,one would expect them to vote with the intent of ensuring the long-term success of the network,which would enable a more streamlined development process as well.Indeed,a

177、 cursory glance at the proposal history of some of the largest Cosmos chains today reveals a prolific pace of protocol upgrades relative to L1s that mostly conduct governance off-chain.At the same time,the use of on-chain governance itself does not necessarily guarantee a productive or risk-free out

178、come.In past reports,weve discussed some of the unintended consequences of the Polkadot networks governance mechanism,which places a heavy emphasis on the Relay Chain as a minimum security threshold.Specifically,the networks reliance on its governance council for a wide range of parameters extending

179、 to its parachains has historically led to a situation where mostly-sovereign parachain teams have been bottlenecked by delays to technical upgrades and approvals from the core development team.In practice,governance in most blockchains,aside from those in the Cosmos ecosystem,is primarily a process

180、 of social consensus that often involves coordination with core developers.Validators and large stakeholders typically have a voice in protocol-level decisions,but these decisions are ultimately carried out by the developers who maintain and release official client software.Such is the case for the

181、Solana network,where validators in the past have played a key role in coordinating network restarts and re-establishing consensus after outages.More recently,the community has begun to establish a more formal governance process in anticipation of future growth.One final note on the importance of gov

182、ernance in blockchains is its role in dictating economic incentives,which is in turn a major factor in economic security.Security in blockchains consist of both technical and economic components;whereas technical security is mostly addressed by the robustness of client codebases,economic risks often

183、 pose a more realistic risk to network stability.Networks that do not feature high economic value bear the risk of well-capitalized actors obtaining a controlling stake.In other words,they have a low cost of attack,and can thus be described as having weak economic security.Many blockchain developmen

184、t teams are funded extensively by the distribution of native tokens from their treasuries,typically managed by a separate foundation entity.These foundations often have significant flexibility in the way they distribute funds,which can theoretically have an impact on both decentralization and securi

185、ty.For example,many blockchain foundations initiated incentive programs throughout 2021-2022 targeted at attracting users and developers,which effectively contributed to a wider distribution of tokens that can be used to secure the network.At the same time,these foundations have also used their trea

186、suries to secure capital for future development,which has the opposite effect of concentrating large portions of token supply within a handful of entities.Though it is not always intuitive,token economics often play a major role in the security and decentralization of blockchain networks.After all,b

187、lockchains fundamentally revolve around the secure,permissionless transfer of money.This dynamic is especially true in the case of blockchains secured by PoS consensus mechanisms,which encompasses nearly all smart contract platforms that exist in the market today.For PoS networks,the value of tokens

188、 that make up the network have a direct impact on the economic incentives afforded to validators,as well as the theoretical cost of attacking the network via consensus.Examples of this interplay between economic value,consensus,security,and decentralization are abundant throughout crypto history,and

189、 they have often revealed the key tradeoffs between various blockchain designs.The Cosmos ecosystem is especially insightful in this regard,given the standard practice of combining consensus and governance among its networks.One notable case involves the Juno network,which,in March 2022,voted to for

190、cibly confiscate the token holdings of a single holder after the core team determined that the holder in question had been able to accumulate far too much of the supply at low 35JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO34cost.This incident highlights the fragility of

191、 PoS networks when the cost of obtaining a controlling stake is relatively low as well as the overriding power of social consensus in extraordinary circumstances.In fact,maintaining or achieving high economic value is a major factor for security and decentralization in most networks that are either

192、new or that have relatively low market caps.Early periods of growth are the most risky for new blockchains,as they essentially face the constant risk of economic attack until it becomes economically unviable.In previous reports,we have commented on the fact that many Cosmos chains feature validator

193、sets that are dominated by only a handful of validators,resulting in concentration of both governance and risk.One way to combat this risk is through the adoption of liquid staking,which essentially allows users to participate in securing the network while maintaining liquid control of their assets.

194、The basic economic security case for liquid staking is that it reduces the opportunity cost of staking tokens to help secure the network.This was a major component of the Cosmos 2.0 proposal introduced in late 2022,which broadly sought to accrue more value and security to the Cosmos Hub,and is part

195、of the reason why liquid staking derivatives(LSDs)have grown to become so popular on Ethereum in recent years as well.When it comes to economic security in multichain systems,one perspective worth noting is the negative impact of fragmented liquidity.As we suggested above,the ease of deploying new b

196、lockchains via the Cosmos SDK goes hand in hand with a period of low economic value and high security risk in the early days of an ecosystem.In theory,this issue would not be as problematic if Cosmos chains all shared the same liquidity and validator set,which would increase the networks overall cos

197、t of attack,as is the case for more integrated L1 chains.This same logic is a major part of the reason why Ethereum,as the second most valuable cryptocurrency by market cap,is often considered a global settlement layer in the broader blockchain ecosystem.2.3 EXECUTIONIn the previous part,we alluded

198、to the fact that modularization and fragmentation of liquidity may pose notable risks in terms of economic security.Low economic value of a network lowers the barrier for a consensus-based attack,and it also makes it easier for the network to become quickly and inadvertently centralized.However,econ

199、omic security alone is not always enough to prevent issues with decentralization.In this section,we explore some of the ways in which transaction execution and its variable underlying parts can grow to become a potentially problematic source of centralization as well.From the standpoint of economic

200、security,the Ethereum network is one of the most difficult to attack among all smart contract platforms,owing to the size of its market cap and the distribution of its nodes.At the same time,its validators and user base are not entirely immune to the allures of economic incentives either.This basic

201、fact can have snowballing implications in terms of decentralization.Remember,validators are responsible not only for providing consensus,but also for transaction execution.One of the main side effects of Ethereum having perhaps the most vibrant on-chain DeFi ecosystem today is that it is also a rich

202、 source of extractable value for opportunistic actors.In recent years,the pursuit of so-called maximal extractable value(MEV)has become one of the most competitive areas of the crypto industry.At a basic level,MEV searchers look for unique opportunities on-chain that allow them to siphon profits fro

203、m organic user activity.This can come in the form of front-running transactions,arbitrage,and more.Often,these strategies are so profitable that it becomes economically viable for MEV searchers to coordinate with validators through bribes.In recent years,this system of competitive bidding for transa

204、ction execution priority has been largely automated through the use of tools like the Flashbots MEV-Boost Relay.This tool has become so prevalent within the Ethereum ecosystem that as of this writing,roughly 33%of all blocks have been proposed by validators running MEV-Boost.From the perspective of

205、users staking ETH for yield,the economically optimal choice often becomes simply choosing validators that capture 37JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO36MEV.Validators have broadly accepted this strategy in order to remain competitive as well.Such tools may arg

206、uably introduce centralization vectors,for example if certain transactions or users are censored by them.However,it must be kept in mind that a larger share can only delay inclusion of censored transactions,but not prevent them(unless a share of 100%is achieved).For example,even if the share of MEV-

207、Boost running validators is one third,it is highly likely that an MEV-Boost censored transaction becomes included in the blockchain in a short amount of time(the probability of inclusion approaches 1 within a minute).The impact of optimizing transaction execution on decentralization extends to broad

208、er efforts to increase blockchain scalability as well,as approaches to interoperability introduce novel challenges.This is apparent when considering the role of communication protocols,which are ingrained in numerous aspects of the blockchain stack.The ability for nodes to communicate is critical in

209、 both vertical and horizontal scaling efforts.In the case of horizontal scaling,as with rollups,multichain protocols or sharding technologies,communication protocols also enable interoperability between often disparate systems.Yet no matter the case,enhancing communication essentially entails reduci

210、ng latency,which often translates into greater hardware requirements.Cross-chain communication protocols exist in various forms today,whether as the Cosmos ecosystems IBC,the Polkadot ecosystems XCM,or even Circles Cross-Chain Transfer Protocol(CCTP)for USDC.In nearly all cases,one of the primary go

211、als of implementing these protocols is to expand the available avenues for liquidity.In theory,the end result of this goal would be a wider distribution of assets across various networks(indirectly increasing decentralization),as well as increasing the overall throughput of blockchain transactions w

212、hen viewed in aggregate.Nonetheless,communication protocols today often suffer from the issue of being non-interoperable with each other.With each new standard for cross-chain communication comes another dimension for centralized security risk as well.Bridge exploits across both L1s and L2s have acc

213、ounted for most of the largest DeFi exploits in history.These risks become especially concerning when they can impact entire networks and vast swaths of capital.For instance,in late 2022,the exploit of the BSC Token Hub revealed a previously unknown bug related to the IBC protocol,highlighting the e

214、xtent of damage that can arise from a single unexpected incident.In recent years,new efforts have emerged to scale blockchains without introducing multiple new layers of complexity and risk,primarily through innovations in execution environments.These approaches typically require a full re-evaluatio

215、n of blockchain architecture from the ground up-which can be challenging on its own-but they also reap the tried-and-true benefits of handling key blockchain functions within an integrated system.In the next section,we take a deep dive into the Solana networks uniquely iterative,execution-centric ap

216、proach toward optimizing scalability and how it stacks up against the most popular smart contract platforms today.39JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO38PART 3SOLANA:THE CASE FOR MONOLITHIC EXECUTIONBroadly speaking,current approaches to addressing the core iss

217、ue of scalability in blockchains can be divided into two main categories:horizontal scaling and vertical scaling.Horizontal scaling generally involves splitting key blockchain functions across multiple systems to alleviate throughput bottlenecks.Multichain protocols such as Cosmos represent an extre

218、me version of this approach,employing multiple app-specific blockchains that can independently perform all the key functions of typical blockchain to distribute network activity.A more nuanced form of horizontal scaling is the use of modular architecture,which involves splitting key blockchain funct

219、ions into separate layers-usually blockchains themselves-that can then be optimized individually to enhance overall throughput.This approach is exemplified by the modern rollup landscape,consisting mostly of various execution layers that settle transactions on Ethereum,as well as specialized data av

220、ailability layers like Celestia.Notably,Ethereum development now follows a rollup-centric roadmap,which began in earnest with the creation of a new PoS consensus layer and subsequent transition from PoW during The Merge.Vertical scaling represents a fundamentally different approach from horizontal s

221、caling,focusing on maximizing blockchain performance within a monolithic architecture.Typically,this means increasing hardware requirements for validators,enabling greater computational capacity and higher transaction throughput while sacrificing decentralization to an extent.However,this is only on

222、e part of the story.As we saw in Part 2,every blockchain function has multiple underlying components,all of which can impact scalability,decentralization,or security during the process of optimization.In this section,we take a focused look at Solanas unique execution-focused architecture,its progres

223、sive evolution as a response to scaling challenges in recent years,and how its scaling approach compares to other major smart contract platforms today.3.1 OPTIMIZING EXECUTION FROM THE GROUND UP“Our premise is that blockchain becomes mainstream.That is only possible if accounts are cheap and useful

224、in all kinds of things,right from finance to games to NFTs.We initially had a goal to be able to smoothly run a billion accounts on chain,to make that possible.we now can do 10 billion at Solanas speed.-Jeff Washington(Solana Labs)41JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEB

225、LOCK.CO40Solana is a monolithic blockchain launched in 2020 with the primary aim of enabling high scalability and fast transaction finality while maintaining low transaction fees.At a high level,Solanas design approach can be described as an unwavering focus on optimizing blockchain execution by con

226、tinually leveraging the state-of-the-art in hardware technology.Central to this approach is the networks use of a custom execution environment,known as the Solana Virtual Machine(SVM).As we touched upon in Part 1,the SVMs Sealevel runtime enables the parallelization of transaction processing,which r

227、esults in an order of magnitude improvement in throughput compared to single-threaded runtimes like the EVM.Solanas parallel processing capabilities are reliant on several other key innovations within its custom blockchain stack that operate in lockstep over the lifecycle of a transaction,ultimately

228、 putting significant strain on validator resources as well.For example,the Solana transaction processing unit(TPU)utilizes each validators kernel space,CPU,and GPU at various stages in transaction processing,depending on whether the validator is running in“leader mode”or“validator mode.”At any given

229、 moment,the entire network will have one validator serving as the leader,which is in charge of producing blocks,ordering transactions within the block,and propagating to all other validators.In validator mode,validators follow a logic scheme known as the transaction validation unit(TVU)-as opposed t

230、o the TPU used by the leader-in order to validate blocks and transmit data sent from the leader node.These processes rely on a proprietary block propagation protocol as well,known as Turbine,so as to not place excessive computational demand on the leader node.The Turbine mechanism requires leaders t

231、o break blocks into much smaller data packets according to the User Datagram Protocol(UDP)standard,which are then retransmitted via validator nodes to others operating in validator mode.In essence,this significantly reduces the bandwidth requirements for leader nodes,with the tradeoff being that UDP

232、 packets limit the size and complexity of transactions sent on the network as well.This,in turn,might mean that a single smart contract/program interaction in the SVM could require far more transactions compared to an equivalent one in the EVM.UDP packets are the base unit for both vote transactions

233、,which refer to transactions sent between validators to achieve consensus,and non-vote transactions,which refer to transactions initiated by users.The outcome of this simplified transaction processing scheme described above is that Solana validators must conduct frequent communication between one an

234、other for both execution and consensus.In fact,vote transactions on the network typically outnumber non-vote transactions by a significant margin,and have been rising steadily over the past year.This messaging-heavy design would naturally lead to excessive latency between Solana validators were it n

235、ot for the networks unique Proof-of-History(PoH)algorithm,which essentially acts as an internal clock that allows validators to be constantly synchronized without additional network communication overhead.PoH is a verifiable delay function(VDF)run by every 43JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROT

236、OCOL PRISMSPONSORED BYTHEBLOCK.CO42validator,which means that validators can always quickly confirm that they are in sync with other validators with respect to this internal clock.This baseline level of additional information effectively replaces some of the typical communication required between va

237、lidators with local computation.Paired with Solanas Tower BFT consensus mechanism,validators on the network are able to reach consensus on hundreds of non-vote transactions per second,representing a major increase in throughput relative to individual EVM chains.3.2 IMPROVING VALIDATOR PERFORMANCE:EX

238、PERIMENTATION IN PRODUCTIONFrom a development perspective,Solanas scaling approach largely entails fine-tuning validator clients(a.k.a.software)to fully leverage the performance of currently available hardware.In principle,this close relationship between standard validator hardware and blockchain pe

239、rformance would translate into increased throughput over time.However,if this logic holds,it would also imply that Solanas validator hardware requirements can be driven higher by both rising network demand and expansions of client responsibilities.Indeed,taking a look at Solana validator documentati

240、on over the years reveals a notable jump in minimum spec requirements between December 2022 and 2023.CPU requirements have remained unchanged since late 2021,but RAM requirements doubled and networking requirements more than tripled over the same period.Interestingly,the NVIDIA CUDA toolkit was requ

241、ired to make use of validator GPUs as late as June 2023,but this requirement has since been removed,suggesting decreased reliance on GPUs going forward.Meanwhile,GPU requirements for RPC nodes specifically increased from 256 GB to 512 GB between December 2022 and 2023.Perhaps the best way to underst

242、and the Solana networks overall scaling approach and current state of development is to trace its evolution over the past few years.As evident from our discussion of Solanas architecture and transaction lifecycle above,the networks liveness and performance rely on a delicate balance between multiple

243、 unique functions carried out by its validators.Any major disturbances to this balance can have a dramatic impact on the networks performance,or,even worse,its security and ability to reach consensus.45JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO44 One of Solanas earlie

244、st major incidents came in September 2021,when the network experienced an outage that lasted about 17 hours and ultimately required a coordinated restart among validators.The cause of the outage was a sudden influx of resource-heavy bot transactions attempting to capitalize on a hotly-anticipated in

245、itial DEX offering(IDO)on the Raydium platform.As explained in the Solana Foundations post-mortem,validators were essentially unable to process the unexpected spike in transactions,which led to an overload of the forwarder queue for unprocessed packets,followed by a rapid increase in proposed forks,

246、which finally caused validator memory resources to become overwhelmed and crash.The September 2021 incident was a sign of things to come,as a string of similar issues would go on to plague the network over the next 1.5 years.Not all incidents led to consensus failures,but a common theme that emerged

247、 was that validators were becoming overwhelmed by sudden spikes in transaction volumes,whether due to hype from token launches and NFT mints,or simply volatility in the markets,ultimately resulting in degraded performance and decreased throughput.Sometimes,these incidents could be directly attribute

248、d to specific software bugs,highlighting the complex challenge of managing validator resources-which are crucial for the Solana network to function properly-under a variety of market conditions.For instance,the network experienced multiple days of partial outages throughout January 2022,ending the m

249、onth with 96%uptime overall.At the time,we noted how one particular bug caused Solana programs to be repeatedly evicted from the software cache,which in turn forced the SVM to recompile these programs,leading to dramatically longer transaction execution times,a.k.a.replay times.Each of these inciden

250、ts helped to identify areas where the Solana validator client was not yet fully optimized to combat large increases in network demand.In January 2022,this took the form of liquidator bots flooding the network amidst the early stages of an upcoming bear market.In late April 2022,bots targeting the Me

251、taplex NFT minting program,Candy Machine,overwhelmed the network to the point of forcing yet another coordinated restart.By then,Solana validators and app developers had begun to implement temporary remedies to combat these issues,such as by penalizing repeated transaction submissions and restrictin

252、g program access to whitelisted accounts during times of abnormally high demand.47JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO46Still,one of the core problems of Solanas design remained largely unaddressed.That is,with trivially low transaction costs and fast confirmati

253、on times,there were almost no economic disincentives preventing bots and malicious actors from repeatedly spamming the network to capture potential market opportunities.These behaviors were further amplified by the fact that,as mentioned earlier,Solana validators use size and complexity-limited UDP

254、packets to propagate blocks,which results in a high baseline number of transactions per second even under normal conditions.After a tumultuous period of network instability throughout H1 2022,Solana developers were finally able to demonstrate notable improvements in validator client performance and

255、resilience after releasing a flurry of key upgrades in June 2022 designed to directly address its historical issues.Chief among these upgrades was the networks initial rollout of the QUIC protocol,which,when integrated with validator TPUs,essentially allows validators to more efficiently filter and

256、sort submitted transactions without significantly impacting throughput.This software implementation would continue to be refined over the next year,but it remains one of the most critical technical upgrades to the Solana network to this day.Solanas issue with spam transactions has also been addresse

257、d with several key features that work more so through economic incentives than software/hardware optimizations.One of these is stake-weighted Quality of Service(QoS),which essentially scales the amount of transactions that can be transmitted from validators to leader nodes based on the size of their

258、 stake in the network.Another major feature that has been implemented is fee markets.Similar to those on Ethereum,Solanas fee markets introduce an economic component to regulate and prioritize transactions,allowing users to submit priority fees to increase the likelihood of their transactions being

259、included in the next block.One major difference is that Solana has implemented local fee markets,which effectively serve to isolate spikes in demand for specific programs without impacting fees and access to other programs on the network.For example,heightened activity surrounding an NFT mint on Sol

260、ana should,to an extent,only increase fees for that particular program.This requires explicit declarations of access to parts of the network state,which would require a significant design overhaul to accomplish in the EVM,and thus represents a feature of transaction execution that is unique to Solan

261、a at the moment.For a more nuanced discussion regarding Solanas design advantages with respect to implementing multi-dimensional fee markets,and the importance of similar functionalities being added to the EVM,see this paper and recent talk by Gauntlet founder Tarun Chitra on the topic.As illustrate

262、d in the chart above showing Solanas network outages,the combination of QUIC,stake-weighted QoS,and fee markets seem to have had a dramatic impact on maintaining network stability.Notably,priority fee usage has continued to rise steadily over the past year as well,suggesting further refinement and a

263、doption of this feature in the future.As of this writing,users of the DEX aggregator,Jupiter,can now set priority features through the web interface,simplifying access to Solanas fee markets for average users.As such,it can be expected that the impact of fee markets on network performance will becom

264、e more pronounced over the next several months.Overall,the timeline of Solanas numerous upgrades over the past year Source:Jupiter49JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO48demonstrates the key challenges involved in scaling a monolithic blockchain,including the op

265、timization of validator performance through computing and networking resource management,as well the balancing of economic incentives for network participants.51JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO50PART 4COMPARING THE CURRENT STATE AND ROADMAPS OF VARIOUS BLOCK

266、CHAIN ARCHITECTURESThe reality is that when it comes to layer ones,there are so many metrics that immediately become absolute nonsense.Once upon a time,TVL was a useful metric and then it very quickly stopped being useful.Because immediately people started gaming it.There is no metric that you can l

267、ook at that allows you to skip thinking.you have to think very clearly about where a number comes from and what it actually means.”Haseeb Qureshi(Dragonfly Capital)Now that we are familiar with the nuances of Solanas architecture and overall development strategy,lets dig into some of the key differe

268、nces between Solanas monolithic scaling approach and modular approaches.We focus specifically on the Cosmos ecosystems multichain scaling approach and Ethereums rollup-centric approach,which represent two of the main manifestations of modular blockchain architecture.Finally,we take an objective look

269、 at the state of blockchain scalability today,and the road ahead from various design perspectives.The current state of each ecosystem along the dimensions of the blockchain trilemma is outlined below.53JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO524.1 THROUGHPUT AND EXE

270、CUTIONFirst,lets take a look at execution,which we established earlier as one of the most important factors dictating blockchain scalability.The Solana teams focus on optimizing transaction processing through engineering has ostensibly given the network an advantage over other blockchains in terms o

271、f throughput;the question is by exactly how much?In terms of raw non-voting transactions per second(TPS),Solana processed a daily average of 239 TPS in October 2023,with a YTD max of 614 TPS reached on March 29th.Over the same period,Ethereum averaged only 12 TPS,reaching a YTD max of 19 TPS on Sept

272、ember 13th.Interestingly,most of the popular rollups today did not feature a higher TPS than Ethereum.In October,zkSync Era,Arbitrum One,and Starknet saw average daily TPS values of 10,7,and 6,respectively.Optimism and Base-which is built on the OP stack-had average daily TPS values of 3 and 6,respe

273、ctively.However,it is important to note that TPS is not accurate as a standalone metric in direct comparisons of throughput between chains,though it may appear so on the surface.For one,TPS is directly affected by transaction volume since it is calculated by simply taking the number of transactions

274、divided by the number of seconds over a certain period.Therefore,a chain with more activity than another at a given time will have a higher reported TPS,even if the two chains have the exact same theoretical throughput.The composition and size of transactions can vary between different chains as wel

275、l.For example,Solana transactions are formatted as UDP packet bundles and have a data limit of 1,232 bytes,whereas Ethereum transactions have a calldata limit of 1 MB as of EIP-4488.A more empirical comparison of throughput between blockchains was conducted by Dragonfly Research(mentioned in Part 1)

276、,which evaluated throughput by assessing how many DEX trades can be fit into a single block.Notably,in this study,the Dragonfly team found that Solana has a theoretical limit of 270 swaps per second on the Orca DEX,achieving similar numbers empirically on the Solana devnet as well.The study,conducte

277、d in late January 2022,assumes average slot times of 590ms on mainnet,and a cost of 74,400 compute units(CU)per Orca swap,given a 48 million CU limit per block and 12 million CU limit per account.As of this writing,the block and account CU limits remain unchanged at 48 million and 12 million,respect

278、ively,but slot times have decreased on average since the original study.For this report,we sought to recreate the Dragonfly study given these updated metrics to gain a sense of how Solana stacks up against other L1s in the present day.Taking a look at epoch 274,roughly when the Dragonfly study was f

279、irst conducted,we can calculate the average slot time with the following formula:Average slot times were 511ms in epoch 274,roughly in line with the figure obtained by the Dragonfly team.We can use the same method to determine average slot times in late 2023,at the time of this writing.Taking epochs

280、 520-524,spanning from October 19 to October 29,we find average epoch slot 55JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO54times of 414ms,419ms,430ms,426ms,and 409ms,for an overall average slot time of 420ms over the five epochs.This alone implies that Solanas throughpu

281、t in terms of DEX swaps per second has increased substantially since early 2022.Another change since the original Dragonfly study is that Orca swaps have become more efficient in terms of CU usage since the protocol implemented concentrated liquidity pools in March 2022.Analyzing a random swap trans

282、action(here)from early 2022 that utilizes the older Orca V2 router reveals a cost of 72,300 CU per swap.Meanwhile,a more recent swap(here)from October 2023 utilizing Orcas V3“whirlpool”router cost only 66,500 CUs,representing an efficiency improvement of 8%.Note that swapping different tokens does n

283、ot result in different CU for a given smart contract version.Therefore,we can base our analysis on a single random datapoint for each,the old and the new version.The same applies for our below analyses for Uniswap and Osmosis.It is important to note that these compute optimizations do not constitute

284、 a direct gain in efficiency for the Solana network itself,but they do have an impact on empirical measurements of throughput when considering DEX swaps as a standard representation of typical network activity.Orcas reduction of CU costs for swaps can also be viewed as an example of the potential fo

285、r further resource optimizations arising from Solana applications themselves that have not yet been fully explored to the same extent as on Ethereum.For instance,looking at a recent Uniswap V3 swap(here)reveals a gas cost of 155,000,essentially unchanged since the Dragonfly study conducted on Uniswa

286、p V2 in 2022,which seems to reflect a tapering of significant protocol-based gas optimization on Ethereum in recent years.Taking into account Solanas 12M per-account CU limits,as well as new values for both slot times(420ms)and swap cost(66,500 CUs),we can extrapolate an updated maximum throughput o

287、f 430 swaps per second.At face value,this already represents an order of magnitude higher throughput than Ethereum(without associated rollups),but it also does not factor in the possibility of throughput gains from parallelizable swaps,which could,for example,originate from non-correlated pairs or l

288、iquidity pools on other DEXs aside from Orca.As a whole,Solanas execution system has a clear advantage over competing networks most of which use the EVM-in terms of transaction throughput.For the sake of completeness,let us also consider performance on Cosmos chains,which were omitted in the Dragonf

289、ly study.Osmosis,like most Cosmos chains,uses the CosmWasm execution environment,and is a suitable example for throughput evaluations given its position as the leader in Cosmos ecosystem DEX volumes.Given a block gas limit of 12.67M,a cost of 430,310 gas per swap(example here),and a 6.03s block time

290、,Osmosis throughput can be determined to be 4.88 swaps per second.Ultimately,the inability of both the EVM and CosmWasm to stack up against the SVMs throughput performance is indicative of the latters technical strengths from an architectural perspective.Rollups and Cosmos chains attempt to address

291、scalability by essentially adding separate channels.While this approach has of course not yielded any major performance enhancements in terms of execution on a single chain,the hope is to increase throughput of the overall ecosystem via adding chains,or layers.In some ways,this situation is the resu

292、lt of deliberate design and tradeoff choices,but some teams that have been primarily focused on offering modularity in the past are now starting to recognize the need for direct execution improvements as well.In a discussion with The Block Pro Research team on execution for Cosmos chains,Osmosis fou

293、nder and original core developer for the Cosmos Hub and IBC,Sunny Agarwal,explained,“Theres a lot of low-hanging fruit in the Cosmos SDK stack,in terms of performance engineering,that was never a top focus for us.We were kind of focused on everything else.”He commented on a recent announcement from

294、the co-founder of Maker-one of the largest DeFi protocols on Ethereum that the protocol would be exploring Solana as one of the options for a template of a new blockchain:That was a bit of a wake-up call,for everyone in the Cosmos ecosystem.It was like,alright,lets just get our act together on that.

295、I think we can spend six months just fixing all these performance issues on the SDK and get it to be within an order of magnitude of the scalability of Solana.I think Solanas scalability comes from a number of sources one,its good engineering,Ill give it to them.They performance engineered it really

296、 well.Certain architectural design choices,some of which I agree with,some of which I disagree with.But the beauty of an app chain is you can decide if those design choices make sense for your app chain or not.For example,Solana does not merkelize state it doesnt have Merkle trees at all.Theres no s

297、uch thing as a light client.And I personally think thats not OK for a public generalized L1 blockchain.But maybe thats OK for an app chain that doesnt care about it,right?So I think the nice thing about the Cosmos idea is like,hey,my chain cares about like client proofs.This chain doesnt.Well,we can

298、 do different things,right?So theres certain architectural things Solana does that,some of which we would adopt in Osmosis,some of which we wouldnt.And then theres the higher node requirements.You have to have pretty beefy servers,which,once again,I think thats an app chain level decision right?Like

299、 is that OK for a generalized L1?In my opinion,no.But is that OK for a perps exchange,maybe.57JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO56Among teams in the Cosmos ecosystem,it is apparent that transaction execution will be an increasingly important priority in the co

300、ming years,especially for a community that has largely focused on enabling developer flexibility in the past.According to Cosmos Hub developer Noam Cohen,one of the key goals for the Hub will be releasing what is known as“Atomic IBC,”which would essentially require merging binaries between some of t

301、he Cosmos chains that are being secured by the Hubs Interchain Security model and enable them to run in parallel.The idea is that Atomic IBC would create atomicity between chains,thus improving composability and helping to scale the current replicated security model.However,Atomic IBC is not expecte

302、d to be released until late 2024 or early 2025.Aggarwal agrees that something like Atomic IBC is probably needed in the long run,but there are also easier targets to focus on in the next year or so,explaining,“Its easier for us to just focus on decreasing block times in Cosmos chains.Today the stand

303、ard is like 5 to 6 seconds.Get this down to one second,sub-second blocks,and then even if you have to do async actions on another chain,its fine.Half a second here,half a second there,I think thats actually a more productive and achievable goal.”In the end,achieving this may require Cosmos validator

304、s to take on a greater computational load than they currently do.As weve seen with Solana,ensuring synchronization between validators through PoH is one of the defining characteristics of the network that enable it to consistently produce sub-second slot times.4.2 HARDWARE REQUIREMENTS AND DECENTRAL

305、IZATIONPart 3.1 breaks down some of the main challenges with vertical scaling through the lens of Solana.Perhaps the clearest takeaway in addition to our discussion in Part 3.2 thus far is that maximizing transaction throughput comes with a major cost:increased hardware requirements.As such,the barr

306、ier to entry becomes much higher for anyone looking to become a validator.This barrier exists in both an economic and technical sense;Solanas validator hardware requirements are not accessible to most people,nor are the skills and time commitment required to keep up with the pace of network developm

307、ents.On a purely technical level,the requirements to run an Ethereum-or Cosmos-node are clearly less stringent than those for a Solana node.As such,the total set of users that could potentially join the Ethereum or Cosmos networks as validators is higher compared to Solana.In theory,Solanas node req

308、uirements thus impact its overall decentralization,at least relative to Ethereum and Cosmos chains as baselines for comparison.However,decentralization itself is difficult to measure objectively.One could simply count the number of active nodes in a network,but this alone is not a useful metric with

309、out knowing how many nodes are controlled by a single entity.Another,somewhat more insightful,metric is the number of nodes required to disrupt a network,commonly known as the Nakamoto Coefficient.The premise behind this metric is that the higher the coefficient,the harder it is to corrupt a critica

310、l mass of nodes,and thus the more“decentralized”a particular network is.Still,the Nakamato Coefficient does not fully account for key architectural variations between different networks,and thus fails to sufficiently capture the nuances involved in assessing decentralization.Even defining the nodes

311、for comparison between blockchains can be tricky.For instance,on Ethereum,anyone can run a node-which enables user interaction with the network-for just the cost of the hardware to do so.However,only validators who stake significant capital(32 ETH)1 are able to propose new blocks.This in turn create

312、s a higher barrier to entry and effectively decreases the number of corrupted entities required to disrupt the network.Another factor to consider is client diversity.Assuming a network uses the same client for all of its validators,this would mean that the cost of censoring the network would effecti

313、vely boil down to a single,centralized point of failure.On Ethereum,this is addressed by having several clients for both consensus and execution.Still,this diversity is best viewed as a spectrum.On the consensus side,Prysm and Lighthouse comprise 77%of all clients,while 85%of execution clients run G

314、eth.Solana co-founder Anatoly Yakovenko has a specific view on decentralization.According to Yakovenko,the real scalability trilemma consists of three 1 Ethereum node operators who use Rocketpool only need to put 8 ETH and 2.4 ETH worth of RPL at stake,theoretically lowering their individual cost.No

315、netheless,these operators still need access to 32 ETH overall,with contributions from users,essentially spreading out the cost burden of staking through Rocketpools coordination mechanism.59JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO58things:1)cost to destroy all repli

316、cas,2)cost to censor a message from arriving to all replicas in real time,and 3)cost of hardware.In his view,Solana has chosen to optimize on 1 and 2,while Ethereum has optimized for 1 and 3.Client diversity is certainly a pending issue for the Solana network today.As of this writing,there are only

317、two major validator clients live on Solana the Solana Labs client,and Jito Labs MEV-optimized fork of the Solana Labs client.This is to be expected in some ways the pace of network upgrades originating through the Solana Labs team over the past year has likely made it practically difficult to develo

318、p an alternate client without significant coordination with Solana Labs.The most significant initiative currently underway to address this issue is the Firedancer validator client,an alternative client being developed by Jump Trading that aims to further maximize throughput on Solana.Yakovenko stres

319、sed the importance of Firedancer in a recent discussion with The Block Pro Research,stating,“I will finally consider Solana to be out of beta when Firedancer launches.”This future may not be all too far off at the Solana Breakpoint conference in late October,the Solana Foundation announced that Fire

320、dancer is now officially live on testnet.There are additional existing efforts to increase the overall decentralization of the Solana network.One of the most significant is the Solana Foundations Delegation Program,which delegates staked SOL controlled by the Foundation to smaller validators if they

321、 are able to meet the comprehensive list of requirements that would indicate reliability over the long term.The Foundation also runs the Server Program,which facilitates easier access to validator and RPC hardware through contracts with data centers around the world.It is important to note that the

322、issue of decentralization can be addressed from a technical standpoint as well.For instance,optimizing resource usage with existing specs is one of the main components of improving scalability.Over time this would lower the cost of running validators(Moores law)and hence allow more people to run a n

323、ode.This was clearly demonstrated with the release of Solanas v1.16 upgrade in September,which appears to have drastically reduced validator memory requirements in the process.Using data collected from the Stakeconomy validator,we can see that average daily memory usage was 132 GB in September,but h

324、as dropped to a daily average 51 GB since the release of v1.16.In other words,increased scalability and throughput does not always translate to higher hardware requirements automatically,especially if optimizations for hardware resources are made on the software client side.For blockchains aside fro

325、m Solana,there is already a non-trivial possibility that validator hardware requirements will eventually need to increase in order to support better execution.According to Osmosiss Aggarwal,it is likely that Cosmos validators will need to expand their responsibilities in the coming months:I think ov

326、er the next six months,youre going to start to see things that cause the SDK to become way more performant.One of the big things were focusing on is having a lot of sidecars.What I mean by that is like having the validators of a chain be able to run secondary things other than the nodes that,for exa

327、mple,basically offload computation.For those nodes,the computation basically has to happen on the state machine,which means every full node has to do it.But if you offload some of it to a validator sidecar and only have validators run that computation,then you can have that submitted as a proof thin

328、gs like that,I think will also be a big focus of both-how do we increase performance,but also how do we decentralize the Cosmos stack?Today for our DEX,what we think a lot about is like,OK,our chain is decentralized,but the front end is centralized,the routing server is centralized,the data indexers

329、 are centralized,and its like-how do we decentralize everything?Well,lets get our validators to run more and more of this.61JANUARY 2024BLOCKCHAIN DESIGN SERIES:PROTOCOL PRISMSPONSORED BYTHEBLOCK.CO60In this age of increasingly complex and demanding user activity on blockchain networks,it is not ent

330、irely far-fetched to expect that validators will soon need to evolve to meet these growing demands.The main question that remains is whether these validator requirements can be kept low enough over time to provide higher performance without drastically sacrificing decentralization or security in bot

331、h the short and long term.4.3 ONWARD AND UPWARD:THE STATE OF BLOCKCHAIN SCALING AND BEYONDIn an overarching sense,the contentious debate regarding the“best”approach to scaling blockchains today boils down to opposing viewpoints regarding the importance of various blockchain properties at different p

332、oints in time.The key problem is that in reality,all of the properties that define a blockchain scalability,decentralization,security,interoperability,usability,etc.should be considered important at any given time.A network that focuses solely on scalability in the short term might quickly become so

333、 centralized that it becomes wholly unable to resist censorship in the future.Similarly,a network that maximizes security at all costs without developing short-term avenues for scaling bears the risk of becoming too expensive or unusable for a new generation of users,threatening its relevance in the long run as well.Meanwhile,any network that doesnt prioritize security at all times must accept the