The promise of an open decentralized internet has been challenged by performance, usability, and energy efficiency problems of the first generation crypto-networks Bitcoin, Ethereum, and their variants. While the development of Ethereum’s new version and its Layer 2 solutions aim to remedy the current performance problems, a new generation of projects Cosmos, Polkadot, and Avalanche have already released extraordinary infrastructures. They aim to scale horizontally with an asynchronous heterogeneous network model, where application-specific blockchains co-exist and interoperate with one another when needed. For inter-chain economic security, they have their own design choices and tradeoffs in their own right, which will have different impact as we will discuss in this article. They aim to create an internet of blockchains to reach a web-scale that can accommodate not just hundreds of thousands (as it is today), but millions of active users per day and realize the user owned and controlled Web 3 vision. This article aims to help developers, researchers, entrepreneurs, investors, and whoever aspires to a decentralized future understand this paradigm shift in crypto-networks.
Topologies of inter-chain economic security in Cosmos, Polkadot, Avalanche.
It is now a common sense that Bitcoin opened Pandora’s Box and became “digital gold” over time. Ethereum introduced programmable internet money and became the platform for crypto-economic innovation. And yet Bitcoin, Ethereum, and their variants have critical issues preventing masses from adopting crypto-networks. We will first look at these issues, then use these points to compare new generation blockchain platforms.
These issues do not only constrain mass adoption of decentralized networks, but push existing users to remain relying on centralized exchanges and hosted wallets. It is too hard for non-technical folks to regularly use a truly decentralized application. On the other hand, existing users continue using Ethereum and Bitcoin because they are not aware of these problems; companies and investors continue to use them because they want to be where liquidity is; and early entrants or “original gangsters” defend these networks because they have large stakes. But another world is possible.
Daily active Ethereum addresses. Source: Etherscan.io
Today, Ethereum accommodates 500,000 daily active users in average, whereas a popular web application like Twitter is used by 200 million daily active users (400x of Ethereum) and Facebook has almost 2 billion daily active users (4000x of Ethereum). Layer 2 and Bitcoin users would add up, but still that is long way away from web scale. Scaling is a highly critical challenge for an open decentralized internet, it is not a problem of tomorrow, but a priority be tackled here and now.
While Ethereum’s new version aims to remedy scaling problems and its interim Layer 2 solutions are currently trying to accommodate the ever increasing demand, a new generation of platforms Cosmos, Polkadot, Avalanche — with mainnets launched in 2019 and 2020 — have reanimated the promise of a true decentralized internet. We will first look into the promises of Ethereum’s new version.
Since its inception, Ethereum’s new version has been changing by adopting mechanisms from new scientific research as well as from inventions of the new blockchain platforms. Ethereum’s new version will use proof-of-stake, split the network into synchronized shards aiming to increase the total computational throughput. Validators running the same Ethereum Virtual Machine (EVM) will be assigned to different network shards, produce blocks, accrue different user activity data, and synchronize with one another from a relay chain called Beacon. However, trying to synchronize all the sharded parts means trying to achieve full replication, that is having consistent copy of the database in all the nodes. This is problematic because the point of sharding in distributed computing is to achieve scaling by not replicating all data in the overall network. In a synchronous model, or a homogeneous network topology, when a shard -for example a popular DeFi shard- accrues way more usage than others, it will start having the same speed, cost, and scaling problems. Plus, there is a new problem of efficiently synchronizing data among shards.
While Ethereum’s transition to new version is said to be fully completed in a year or so, what are known as Layer 2 solutions — rollups (Optimistic, zkSync), plasma, and state channels — have launched to provide efficiency and speed for the increasing demand to use Ethereum. The dilemma is that Layer 2 trust models have either intermediary central operators defeating the purpose of decentralization and censorship-resistance, or multiple incentivized operators (i.e. Polygon is built with Tendermint and running on multiple validators, Matter Labs is aiming for a validator network with zkSync), which is akin to being another decentralized blockchain with its own token (e.g., MATIC), and eventually competing with its Layer 1. Hence, these single chain architectures will hit the same transaction cost issue as more users join them.
Most recently Ethereum adopted a new strategy called rollup-centric roadmap, which positions Ethereum Layer 1 for data availability and Layer 2 projects for computation. In other words, Ethereum wants to become the base layer to guarantee data availability and share security with rollups. As a result, Ethereum is embracing an ecosystem of EVM blockchains for computation, whether a single rollup dominates or multiple rollups co-exist (see Vitalik Buterin’s article Endgame). In fact, this strategy fits into the emerging modular blockchain design, where a blockchain can outsource data availability or execution to other blockchains. A generalized model for this strategy is developed by Celestia and EigenLayr. Moreover, Ethereum’s new strategy is resembling the shared security models already used in Polkadot and Avalanche.
On the other hand, because Cosmos, Polkadot, Avalanche have bridges to Ethereum on at least one of their EVM compatible chains, they are sometimes put in the same “Layer 2” bucket, whereas these projects often call themselves Layer 0, as they provide infrastructure for building interconnected Layer 1 blockchains.
Cosmos, Polkadot, Avalanche aim to scale horizontally with an asynchronous heterogeneous network model, where application-specific blockchains have different virtual machines and can interoperate with other chains when needed. These infrastructure platforms provide capacity to build your own custom blockchains, which enables a much larger design space for decentralized applications and assets. Running your project as a sovereign chain instead of a set of smart contracts has three fundamental advantages:
These new generation networks also have bridges to Ethereum and soon to Bitcoin, as well as bridges to one another are under development to fully realize the internet of blockchains vision.
Cosmos, Polkadot, Avalanche have critical differences at the protocol level (e.g., consensus mechanism, economic security topology) that impact platform features (e.g., inter-chain communications, token economics, types of possible applications) and how they scale their networks (e.g., validator participation, staking attributions). The comparison below aims to help developers, entrepreneurs, investors, researchers, and those who consider building on these new generation infrastructures understand the differences between these architectures and their trade-offs.
Secure and consistent replication of an application state on an open network of machines is achieved by a consensus mechanism. While doing this, the network should remain fault tolerant with valid consensus despite imperfect information or malicious actors (Byzantine Fault Tolerance). The Practical Byzantine Fault Tolerance (PBFT), used in Cosmos and and Polkadot, requires all participating nodes to talk to each another, so the network agrees on a decision with absolute certainty. It has low latency and quick finality, but it cannot scale to many participants in a global open network because load on each validator node increases exponentially as the validation work increases. Bitcoin introduced the longest chain consensus mechanism (Nakamoto Consensus) that allows probabilistic finality and astronomically low error rate. It allows a robust and scalable network over time, but it is very slow.
Allowing participation in the consensus of an open network while preventing operation of multiple identities by the same entity (sybil attack) is handled by a proof-of-work or a proof-of-stake mechanism. Like all new projects, Cosmos, Polkadot, Avalanche use proof-of-stake due to its energy efficiency and its capacity to provide larger design space. There are also projects on these networks that implement a lighter proof-of-work mechanism for fair coin distribution mechanism.
The total amount of computational work per second a network can handle depends on the complexity of the virtual machine and runtime functions used on the network. Cosmos, Polkadot, and Avalanche are building specialized asynchronous blockchain networks, so eventually their network as a whole is unbounded in terms of throughput. What really matters is how much these networks can grow, and their choices of inter-chain economic security matter.
As the activity grows on the overall network, transaction fees increase. Cosmos, Polkadot, Avalanche build specialized networks where each chain has its own custom fee mechanism based on their own state growth.
The numbers below are from March 17, 2022.
Decentralization is also a function of validator stake and reward concentration (rewards weighted based on stake), which typically follows a long-tail distribution –few validators have the most stake while many validators have few. Fair stake distribution is still an open problem for blockchain platforms and each project is trying to achieve fairness differently. For example, Polkadot can have a limited active validator set due to its PBFT-based consensus at its core, but those active validators are rewarded equally through the Phragmén election method. Avalanche can have unbounded number of active validators due its novel consensus mechanism, and average validator weight is dropping progressively, increasing its level of decentralization.
The numbers below are from March 17, 2022.
Bridging between blockchains with separate security levels without some sort of a security sharing mechanism, as in current Cosmos architecture, is not that different than bridging any chain in general. So without common finality guarantees, inter-chain communication has varying risk levels. Polkadot’s inherited security model allows unified finality guarantee, and under that umbrella parachains can pass arbitrary data to one another safely. Avalanche’s overlapping network of validators model currently enables security sharing between chains in the primary network, and soon between the chains in different subnets directly without a need for a bridge. So, the more subnets have overlapping validators (who have shared interests in both subnets), the higher security guarantees their communication can have. In general, overlapping validators between different chains (like merged mining in proof-of-work) can provide more secure inter-chain communication.
All blockchains in their core have the following components: a database, a p2p network, a consensus mechanism, a transaction handling mechanism, and state transition functions (runtime or virtual machine). Cosmos, Polkadot, Avalanche provide these core components and let developers build their custom state transition functions.
Hosting web-scale user activity has a better chance via asynchronous network of specialized blockchains than a network of blockchains running instances of the same virtual machine (i.e., Ethereum’s new version). In this section, we discuss how blockchain networks and inter-chain communications are composed in more detail for Cosmos, Polkadot, and Avalanche.
The Cosmos ecosystem has a distributed network topology, where different blockchains with different purposes have their own validator set, and these chains communicate to one another via bridges when needed. This topology is criticized to be as secure as the least secure chain (when the most secured chain accepts assets from the least secured chain, it becomes less secure). However, it also makes the overall network resilient that there is no one individual chain whose security is critical to the entire ecosystem’s survival. But then, how is the Cosmos ecosystem different than pretty much any blockchain bridging to other chains? Cosmos has the “no strings attached” policy, which allows projects such as Binance DEX, Oasis, Terra, Nym, and many others to use Tendermint to develop and launch their own application specific blockchains.
The Inter-Blockchain Communication (IBC) protocol connects the blockchains in the Cosmos ecosystem (see the 28 interconnected chains on Map of Zones). As chains implement the IBC protocol they bridge to one another and the overall Cosmos ecosystem liquidity increases. IBC follows pretty much how blockchain bridges work. When you send an asset from one chain to another, i) you lock them in the source chain, ii) then a third party -possibly federated- relayer, who is monitoring the chains, picks up the receipt and delivers to destination chain, iii) the receiving chain verifies the receipt and gives you a representation of assets in the source chain. Within the Cosmos ecosystem, chains that implemented IBC have the Tendermint light client verifiers so that they can consume and verify those receipts in their communication. Also, IBC is a general protocol that can be implemented in different blockchain architectures (there is an IBC implementation for Substrate). Furthermore, the new IBC versions will have shared security schemes (see Billy Rennekamps’s talk for more).
Polkadot has a hierarchically inherited security topology, which is efficient for arbitrary data communication between its parallel chains (parachains), yet those parachains are dependent on leasing security from a central relay chain. Polkadot parachains do not need to build a validator community, instead they lease security from the relay chain. They do this by winning a slot (total ~100 slots) in an auction and locking Polkadot’s DOT tokens (they raise DOT funds via crowdloans). When these domain-specific parachains get connected and synchronized to the relay chain via their collator nodes, their functionalities become immediately available. One criticism of this mechanism is that different chains may not need the same level of security, and furthermore, that there shouldn’t be any single chain whose security is critical to the ecosystem’s survival. Although Polkadot’s narrative promotes the idea of parachains without validators today, one can just start a blockchain using Substrate and grow a validator community without relying on the central relay chain (see Compound Gateway). Furthermore, a parachain can grow its own validator community, unlock its DOT funds at the end of the lease period, and use bridges when cross-chain communication is needed. Moreover, there can be multiple relay chains which benefits the overall Polkadot ecosystem. The hierarchical topology will most likely remain because inherited-security-enabled cross-chain communication is more efficient than using bridges between parachains.
Polkadot developed Cross-Consensus Message Format (XCM), a generalized format for communication not only between parachains, but also between different smart contracts, bridges, and Substrate pallets. XCM works with Vertical Message Passing (VMP), which enables message exchange from the relay chain to parachains and back, and Cross-Chain Message Passing (XCMP), which allows the parachains to exchange messages with other parachains on the same relay chain. A message in XCM is a program that runs on Cross-Consensus Virtual Machine (XCVM) (see Gavin Wood’s article series). This abstraction for programing networks and building composable inter-chain applications can be also used in other heterogeneous blockchain networks.
As the parachains grow their community, they may also want to have their own validator sets (see Acala’s presentations), so that they can become a relay chain leasing security to other chains. Although a nested security sharing mechanism can get complex, all sub-parachains would share common finality guarantees, and the total number of state transitions per second would increase, expanding the aggregate computational throughput of the entire Polkadot network.
Avalanche has an overlapping networks topology. Every Avalanche validator node has to secure the Primary Network while securing other subnetworks. A set of validators form a subnet. A subnet can validate multiple blockchains, whereas each blockchain is validated by exactly one subnet. In other words, a validator node may be a member of many subnets. When you start a new chain you have to provide incentives to attract a subnet of validators, who already run the Primary Network and may possibly be running other chains. If your chain is attracting new validators, then they have to be able to run the Primary Network as well as the subnet that runs your chain. Overall, the subnet architecture enables overlapping network of validators (see the diagram above), which stems from the novel Avalanche consensus mechanism. Because Avalanche consensus does repeated sub-sampling among its validator nodes, it needs not all nodes but a small set of nodes to communicate to one another, which results in low messaging complexity in the network. So, the bandwidth and processing power requirements per node stays constant even as the network grows to many thousands of validators. As a result, in terms of validator participation, chains built on Avalanche are more inclusive than Polkadot and Cosmos because of its unbounded participation per chain. How many chains a validator can run depends on the chain runtime / virtual machine design complexity and currently remains an open question.
Interoperability between Avalanche chains is efficient not only because of the fast finality but also being in the same Primary Network enables sharing common finality guarantees (currently asset transfer between X-Chain, P-Chain, and C-Chain are near instant). Security sharing model is different than Polkadot or how it is envisioned in new Ethereum rollup-centric ecosystem. Avalanche’s novel subnet architecture enables higher density networks. This is because security sharing takes place not only between the chains in the Primary Network, but among chains in all overlapping subnets. This allows composability and programmability of networks, opening a new design space, while enabling a type of group-forming network (see Reed’s law) that can scale exponentially to many millions of daily active users to materialize the Web 3 vision.
Heterogeneous blockchain networks Cosmos, Polkadot, Avalanche offer a vast design space with their core infrastructure innovations. To this day, Ethereum has been the place where crypto-economy innovation has taken place. In fact, teams building on these new networks initially created glorified versions of what exists on Ethereum (decentralized exchanges, automated market makers (AMMs), lending, stablecoins, aggregators, insurance, NFT platforms etc.), but there are also projects discovering novel use cases by taking advantage of these new infrastructures.
On the Cosmos network, Osmosis combines transaction privacy (using threshold decrypted transactions to prevent front-running) with cross-chain AMM functionality and has IBC implemented for bridging with other chains. Celestia encodes block data to improve security of light clients, which is a key component for chain interoperability between self-sovereign chains and their varying security levels in a distributed chain ecosystem. Regen enables a crypto-economic platform to incentivize regenerative agriculture and utilizes data from sensors and satellites with an auditor ecosystem. Nym enables mixnet which prevents network traffic analysis from an adversary capable of watching the entire network. Nym uses Tendermint and Cosmwasm smart contract controlling the directory service, node bonding, and delegated mixnet staking. Penumbra enables privacy-preserving cross-chain network transactions. Tendermint is also used by large projects such as Binance DEX and Terra. Even greater value will be unlocked when these separate blockchain networks start to interoperate via IBC.
On the Polkadot network, Acala parachain is a DeFi hub that provides functionality from AMM to lending to stablecoins. Moonbeam is an EVM-compatible smart contract chain. Subsocial is building a decentralized social network platform. Robonomics is building autonomous robot services. Bit Country is a platform for launching your virtual world / metaverse for your community. Integritee and Phala use Trusted Execution Environments (TEEs) to enable decentralized confidential computing and encrypted data storage. Polkadot’s development framework Substrate is also used independently (not as parachain) for running blockchains such as Compound Gateway. While all parachains are compatible with Polkadot’s inter-chain ecosystem by design, they should really be harnessing the incredible composability, memory-efficiency, and auto-upgrading meta-protocol governance capabilities of the Substrate framework to enable novel use cases.
Avalanche’s EVM compatible C-Chain initially attracted developers building efficient versions of Ethereum projects. Pangolin is a fast AMM cloned from Uniswap. Sherpa Cash enables private transactions, cloned from Tornado. TraderJoe started as an AMM and added lending on its way to becoming a DeFi hub. Benqi lend-borrow application is a version of Compound, but also started liquid staking for AVAX. Platypus is a better version of Curve stable-swap in that it has asset liability management. The largest Ethereum projects such as Aave, Curve, Sushiswap who employ a multi-chain strategy also launched on C-Chain and attracted massive liquidity, which are being fueled by the thriving Avalanche-Ethereum Bridge. Avalanche ecosystem also has new type of assets, one of which is for litigation financing, when combined with DAOs, it may have a huge impact in bridging existing legal systems with crypto-networks. In fact, the ingenious Avalanche consensus and the overlapping subnets topology together provide a tremendous innovation space for new projects to come.
Heterogeneous blockchain networks Cosmos, Polkadot, Avalanche provide extraordinary infrastructures to enable internet of blockchains, which demonstrates that asynchronous heterogeneous network model works efficiently and it is an improvement over Bitcoin and Ethereum as they are now. They will eventually accommodate millions of daily active users, and achieve the user owned and controlled Web 3 vision.
Co-existence of these major architectures is healthy for a true decentralized internet, as they have their own design choices and tradeoffs in their own right. Understanding the differences and similarities of these new infrastructures today would help building future-proof systems. Projects using these infrastructures will go beyond smart contract applications, become scalable production quality systems with their own specialized chains and communities, and demonstrate previously unimaginable use cases. As this might happen in a vacuum, we are still left with open questions: How might we make sure liquidity efficiently flows between inter-chains instead of being siloed within particular chains? How will those open organizations running across chains prevent emergence of multi-chain whales and ensure fair distribution of wealth and power?
[1] See Bitcoin’s Academic Pedigree paper by Arvind Narayanan and Jeremy Clark for several decades’ worth of cryptography research Bitcoin is built on.
Special thanks to Sam Hart, İstem D. Akalp, Engin Erdogan, Joe Petrowski for their feedback and review.
Disclosure: The author may hold positions in the assets of the projects discussed in this article.