Cosmos — an early in-depth analysis at the ecosystem of connected blockchains — Part One
This is part one of three articles where i will discuss what i have learnt whilst looking into Cosmos. I will provide links throughout the article to provide reference to sections as well as a list of sources at the bottom of the article for you to look into specific areas in more detail if required. Hopefully it will be useful for those interested in learning more about the project.
Cosmos is still very early in development process with components such as IBC which connects two blockchains together currently in research / specification stage, as a result can change by the time its released.
What is Cosmos?
Cosmos is a network and a framework for interoperability between blockchains. The zones are powered by Tendermint Core, which provides a high-performance, consistent, secure PBFT-like consensus engine, where strict fork-accountability guarantees hold over the behaviour of malicious actors. Cosmos is not a product but an ecosystem built on a set of modular, adaptable and interchangeable tools.
In Tendermint, consensus nodes go through a multi-round voting proposal process first before coming to consensus on the contents of a block. When 2/3 of those nodes decide on a block, then they run it through the state transition logic providing instant finality. In current proof of work consensus for Ethereum, the consensus process is inverted, where miners pick the transactions to include in a block, run state updates, then do “work” to try and mine the block.
Tendermint BFT can handle up to thousands of transactions per second (depending on the number of validators). However, this only takes into account the consensus part, the application layer is the limiting factor though. Ethermint (described below) has achieved up to 200 tps to give you an idea of the speed available per blockchain which is significantly more than current versions of Ethereum and Bitcoin etc.
The Tendermint consensus is used in a wide variety of projects, some of the most notable include Binance Chain, Hyperledger Burrow. It’s important to note though that just using Tendermint consensus doesn’t mean they can connect to other chains with the cosmos ecosystem, they would need to fork their code to implement IBC as a native protocol to allow interoperability through IBC.
The Tendermint consensus algorithm follows a traditional approach which relies on all validators to communicate with one another to reach consensus. Because of the communication overhead, it does not scale to 1000s of validators like Bitcoin or Ethereum, which can have an unlimited number of validators. Tendermint works when there are 100s of validators. (Cosmos Hub currently has a maximum of 100 validators and the maximum tested so far with Tendermint is 180 validators)
Therefore, one of the downsides of a blockchain built using Tendermint is that, unlike Bitcoin or Ethereum, it requires the validators to be known ahead of time and doesn’t allow for miners to come and go as they please.
Besides this, it also requires the system to maintain some notion of time, which is known to be a complex problem in theory. Although in practice, Tendermint has proven this can be done reasonably well if you use the timestamp aggregates of each node.
In this regard, one could argue that Tendermint consensus protocol is “less decentralized” than Bitcoin because there are fewer validators, and they must be known ahead of time.
Tendermint’s protocol guarantees safety and liveness, assuming more than 2/3 of the validators’ voting power is not Byzantine (i.e., malicious). In other words, if less than 1/3 of the network voting power is Byzantine, the protocol can guarantee safety and liveness (i.e., validators will never commit conflicting blocks at the same height and the blockchain continues to make progress).
Cosmos goal is to provide sovereignty through governance to developers by making it easy to build blockchains via the Cosmos SDK and provide interoperability between them, using Tendermint consensus. This is their main differentiator compared to competition like Polkadot and Ethereum 2.0. Ethereum 2.0 and Polkadot are taking a different approach by only using shared security, where there is a root chain which controls the security / prevents double spending for all connected blockchains.
Governance is where all stakers vote on proposals to determine what changes are implemented in the future for their own blockchain, stakers can either choose to delegate their vote to the validator or they can instead vote directly. Without sovereignty all DAPPs share the same underlying environment. If an application requires a new feature in the EVM it has to rely entirely on the governance of the Ethereum Platform to accept it for example. However, there are also tradeoffs to having sovereignty as each zone is going to need a way to incentivise others to validate / create blocks on the Zone by running Full Nodes. Whilst it may be easy to create a blockchain using the cosmos SDK and to mint a token, there are the legal costs / regulation associated with creating your own token. How are you going to distribute the tokens? How are you going to list them on exchanges? How are you going to incentivise others to use the token without being classed as a security? All of which have led to a significant reduction in the number of ICOs being done. With every zone needing their own validator set, there’s going to be a huge number of validators required each trying to persuade them to validate their zone with only a finite number of validators available.
Each Zone / App is essentially a mini DAO and not all are going to be comfortable about having their project progress been taken out of their hands and instead relying on the community to best decide on the future (unless they control 2/3 of the tokens). The Cosmos Hub has proved this can be successful, but others may be risk averse to having their application be a mini DAO. Should someone / competitor acquire 1/3 of the tokens of a zone then they could potentially prevent any further progress being made by rejecting all governance votes (this would be very costly to do on the Cosmos Hub due to its high amount staked, but for all the other less secure zones this potentially may be an issue).
Security for some zones will likely be a lot lower with every developer needing to validate their own blockchain and tokenise them with POS with no easy way to validate the setup of a validator to ensure its secure. Whilst the Cosmos hub is very secure with its current value staked, how secure zone’s will be with significantly less staked remains to be seen. Whilst providing soverignty was Cosmos’s main goal from the start, they are also looking at being able to provide shared security by having validators of a connected Hub also validate /create new blocks on the connected zone’s blockchain for them as well. They are still going to need some way to incentivise the validators to this. Another option is if the developers didn’t want to create a token, nor want sovereignty etc, then they could just build a DAPP on the EVM on a zone such as Ethermint.
As can be seen their are potential advantages and disadvantages to each method, but rather than forcing shared security like Ethereum and Polkadot, Cosmos is giving the developer the choice so will be interesting to see which they prefer to go for.
Layers of a blockchain
From an architecture standpoint, each blockchain can be divided into three conceptual layers:
- Application: Responsible for updating the state given a set of transactions, i.e. processing transactions.
- Networking: Responsible for the propagation of transactions and consensus-related messages.
- Consensus: Enables nodes to agree on the current state of the system.
The state machine is the same as the application layer. It defines the state of the application and the state-transition functions. The other layers are responsible for replicating the state machine on all the nodes that connect to the network.
The Cosmos SDK is a generalized framework that simplifies the process of building secure blockchain applications on top of Tendermint BFT. The goal of the Cosmos SDK is to create an ecosystem of modules that allows developers to easily spin up application-specific blockchains without having to code each bit of functionality of their application from scratch. Anyone can create a module for the Cosmos SDK and using ready built modules in your blockchain is as simple as importing them into your application.
The Tendermint BFT engine is connected to the application by a socket protocol called the Application Blockchain Interface (ABCI). This protocol can be wrapped in any programming language, making it possible for developers to choose a language that fits their needs.
Hub and Spoke Topology
Cosmos follows a hub and spoke topology as its not feasible to connect every zone together. If you were to connect every blockchain together the number of connections in the network would grow quadratically with the number of zones. So, if there are 100 zones in the network then that would equal 4950 connections.
Zones are regular heterogenous blockchains and Hubs are blockchains specifically designed to connect Zones together. When a Zone creates an IBC connection with a Hub, it can automatically access (i.e. send to and receive from) every other Zone that is connected to it. As a result, each Zone only needs to establish a limited number of connections with a restricted set of Hubs. Hubs also prevent double spending among Zones. This means that when a Zone receives a token from a Hub, it only needs to trust the origin Zone of this token and each of the Hubs in its path. Hubs do not verify or execute transactions committed on other zones, so it is the responsibility of users to send tokens to zones that they trust.
There will be many Hubs within Cosmos network the first Hub to launch was the Cosmos Hub whose native staking token is called ATOM. ATOM tokens are specific to just the Cosmos Hub which is one hub of many, each with their own token. Transaction fees for the Cosmos Hub will be payable in multiple tokens so not just ATOMs whereas other Hubs such as IRIS has made it so that all transaction fees are paid in IRIS for transactions on its hub.
As mentioned, the Cosmos Hub is one of many hubs in the network and currently has a staking ratio of around 70% with its token ATOM having a market cap of just over $800 million. IRISnet was the second Hub to launch which currently has around 28% bonded with its token IRIS which has a market cap of just under $17 million. The Third Hub about to be launched later this month has its token SENT which has a market cap of around $3.4 million. As you can see the security of these 3 hubs differ wildly and as more and more hubs and then zones are brought online there is going to need to be a lot of tokens / incentivisation for validators.
Standard Cosmos zones / hubs don’t have smart contract functionality and so to enable this, as the Application layer is abstracted from the consensus layer via ABCI API described earlier, it allows Cosmos to port the code over from other blockchains such as Ethereum and use it with the Tendermint Consensus to provide access to the Ethereum Virtual Machine. This is what is called Ethermint.
This allows developers to connect their zones to specialised zones such as Ethermint to build and run smart contracts based on Solidity, whilst benefiting from the faster performance of the tendermint Conensus over the existing POW implementation currently. Whereas a normal Go Ethereum process runs at ~12.5 transactions per second (TPS), Ethermint caps out at 200 TPS. This is a comparison against existing Ethereum speeds, whilst obviously Ethereum are working on their own scaling solutions with Ethereum 2.0 which will likely be ready around the same time. Existing tools / dapps used on ethereum should easily be able to be ported over to Ethermint by the developer if required.
In addition to vertical scaling (with the increase in tps by using Tendermint consensus), it can also have multiple parallel chains running the same application and operated by a common validator set. So if 1 Ethermint zone caps out at 200 TPS then 4 Ethermint zones running in parallel would theoretically cap out at 800 TPS for example.
There is a huge number of developers / apps currently built on Ethereum, should a developer choose to migrate their DAPP over to Ethermint they would lose native compatibility with those on Ethereum (except through Peg Zone), but would gain compatibility with those running on Ethermint and others in the cosmos ecosystem.
IBC stands for inter-blockchain communication protocol and is an end-to-end, connection-oriented, stateful protocol for reliable, ordered, authenticated communication between modules on separate distributed ledgers. Ledgers hosting IBC must provide a certain set of functions for consensus transcript verification and cryptographic commitment proof generation, and IBC packet relayers (off-chain processes) are expected to have access to network protocols and physical datalinks as required to read the state of one ledger and submit data to another.
In the IBC architecture, modules are not directly sending messages to each other over networking infrastructure, but rather creating messages to be sent which are then physically relayed via “Relayers”. “Relayers” run off-chain and continuously scan the state of each ledger via a light client connected to each of the 2 chains and can also execute transactions on another ledger when outgoing datagrams have been committed. For correct operation and progress in a connection between two ledgers, IBC requires only that at least one correct and live relayer process exists which can relay between the ledgers. Relays will need to be incentivised to perform this task (the method to which hasn’t been established as of this writing)
The relay process must have access to accounts on both chains with sufficient balance to pay for transaction fees. Relayers may employ application-level methods to recoup these fees, such by including a small payment to themselves in the packet data. More information on Relayers can be found here
A high-level overview of the process is that Zone 1 commits an outbound message on its blockchan about sending say 1 x Token A to Hub1 and puts 1 x Token A in escrow. Consensus is reached in Zone 1, and then it’s passed to the IBC module to create a packet which contains the reference to the committed block, source and destination channel/ connection and timeout details and is added to Zone 1’s outbound queue as proof.
All relayers (who run off-chain) are continuously monitoring the state of Zone 1 via the Zone 1 light client. A Relayer such as Relayer 1 is chosen and submits a proof to Hub1 that Zone 1.
Hub 1 then sends a receipt as proof that it has received the message from Zone 1, relayer1 sends it to Zone 1. Zone 1 then removes it from its outbound queue and sends proof via another receipt to Hub1. Hub1 verifies the proof and mints the token.
This video below explains the process in more detail as well as covers some of the other points i raise later in this article so worth a watch (time stamped from 22:24 to 32:25) and also here from 38:53 to 42:50
Whilst there is an option for UDP style transfer where a zone will send a message to a Hub and it doesn’t care whether it gets there or in any order etc, Token transfers are going to require the TCP style connections in IBC where there is a send, receipt and then another receipt as explained above. Each Send, receipt followed by another receipt is going to take at least 2 blocks and so using Cosmos Hub block times as an example with 6.88 second block times a transfer between one zone and hub could take a minimum of 41.28 seconds. You also then have to factor in the amount of other transactions going through those at that time and relevant gas price to see whether it is able to use 2 consecutive blocks or whether it may take more. This is also explained in this video “ILP Summit 2019 | Cosmos and Interledger | Sunny Aggarwal” (time stamped) from to 12:50 to 15:45
In Part Two we will look at potential issues with multi hop routing, token transfers across multiple routes and Peg Zones, whilst also looking at other interoperability solutions that would resolve some of these issues and compliment the cosmos ecosystem. Part Two can be found here
Intro to Cosmos, Whitepaper, Token Distribution Details, Diagram showing how Tendermint works, Tendermint details, List of Projects using Tendermint, Cosmos Validators FAQ, Peg Zones, Ethermint, IBC, IBC BI-Weekly Calls, Cosmos SDK, Relayers, ATOM Tokenomics
Cosmos Explorers , Validator info — Mintscan, B-Harvest, Link to lots of other sites for wallets, delegating etc, Number of ATOMs
Other recent articles:
Blockchain Brad Interview with Quant Network CEO — 570 Banks, Rearchitecting the Financial System (DeFi), QNT being listed on Traditional Stock Exchanges this year, Expanding to Asia, Partnered with Oracle, Connecting Binance Chain to Enterprises, Telecoms, Patents and much more