The "Modular" Era of Ethereum: PoS, Sharding, and Rollups

Original title: "Ultra Scalable Ethereum"

Original author: David Hoffman, Bankless

Original translation: Nanfeng, Unitimes

Ethereum development is reaching a new level of maturity. The gap between where Ethereum is now and its defined roadmap is rapidly narrowing.

We are now at a stage where it is clear that Ethereum is developing a modular design architecture. The properties that make a blockchain a "blockchain" are showing a trend of differentiation to enable each chain to achieve its own optimal vision.

In this article, we will explore how PoS (Proof of Stake Consensus Mechanism), sharding, and Rollups will achieve modular blockchain design, thereby realizing Ethereum's long-term vision and setting standards for the future development of blockchain.

01. The “Trilemma” of Blockchain

The infamous “Blockchain Trilemma” (also known as the “Impossible Triangle” of Blockchain) states that you can only optimize two of the three properties of a blockchain, and one of them must be sacrificed due to technical limitations. These three properties are:

· Scalability:What is the data throughput of the system? How big is the TPS?

· Decentralization:How many nodes are there in the network? Are there any power centers?

· Security:How difficult is it to attack the network?

Impossible Triangle of Blockchain: Each side of this triangle can only achieve two properties

Why is this the case? Why can't blockchain achieve sufficient decentralization, security and scalability at the same time?

This is because traditional blockchains are monolithic. Blockchains always try to achieve all three goals on the main chain (L1). However, when we modularize these components, the limitations of the blockchain trilemma disappear.

For an analogy, think of the division of labor. This economic principle divides a complex task into smaller parts, each of which can specialize in a specific part of the work, so that the output of the whole system far exceeds that of the same number of workers working alone.

So what does a modular blockchain look like? How does it work? Before explaining this, we need to understand the three components of a blockchain, which make up the three major properties of a blockchain described above.

Blockchain's Underlying Components

Decentralization, scalability, and security are the three main properties of blockchain. They are features that a blockchain can include, but there also needs to be underlying components that implement these three properties. Modular blockchains divide these components into separate parts and maximize them. So what are these underlying components?

· Consensus: Provides security for the blockchain and defines the authoritative authenticity of the data stored on the blockchain. The consensus of the blockchain determines what the current block number is and what is contained in "block N". (Note: N represents the block number)

· Execution: The calculations required to update the blockchain from "block N" to "block N+1". That is, taking the old state, adding some new transactions, and then converting the old state to the new state. This process needs to be done through calculation.

· Data availability: Data that can be referenced is guaranteed by the blockchain L1 (main chain). All the data that constitutes block N.

Before we go into the explanation, let's use an analogy to familiarize ourselves with these terms. Suppose it is a Wednesday morning and you are heading to a local Wells Fargo branch to deposit a $100 check.

· Your account status is your bank balance, which currently has $69,420.

· All account transactions from the time you opened the account to the present are stored in the Data Availability Layer, a centralized database hosted and protected by Wells Fargo.

· When the teller at this branch processes your check, Wells Fargo actually performs a state transition on the Data Availability Layer, updating your account balance to $69,520.

· This new “N+1” state ($69,520) is now reflected in your Wells Fargo mobile app, web, and other branches. This consensus is possible because all updates occur in a centralized database that only people with credentials can access.

Now, let’s extend this example to the blockchain:

Consensus

As mentioned earlier, consensus defines the authoritative authenticity of the data stored on the blockchain.

In the category of consensus, there are two main types of consensus: PoW (Proof of Work) and PoS (Proof of Stake). These consensus mechanisms determine how new blocks are added to the blockchain and how consensus participants reach agreement on the correctness of the blocks.

By reaching consensus, the blockchain can move forward in time without being split into millions of different chains with their own versions of the authenticity of the data.

· Execution

The execution property of a blockchain involves carrying the state of the blockchain to the next new block.

Block N contains some specific states that represent how data (such as state balances, contract code, etc.) has been transformed from block N-1. The validator then grabs some new transactions from the blockchain's mempool (also known as the transaction pool) and updates the state in block N by producing block N+1, that is, getting the state of block N and then changing this state based on the transactions grabbed from the memory pool. The blockchain's memory pool (mempool) contains all the transactions waiting to be packaged and confirmed by miners/validators, just like a group of customers in a bank waiting to get banking services.

When validators use transactions selected from the memory pool (as input) and the consensus mechanism to calculate the next state of the new block, it means that these transactions are executed.

· Data availability

Data availability refers to the data hosted on each blockchain node. If the data exists on these nodes, then the data is available to anyone using the blockchain.

This also makes this data very important and expensive. The amount of data available is very limited (we call it scarce block space). When you add some data to the blockchain (such as sending a transaction or deploying a contract), you are actually adding the data to all computers (nodes) running nodes of this blockchain, and it is permanently added. As we know, the records on the blockchain are immutable, which means that the data stored in the system can be said to be the most valuable data created by humans.

Everyone wants to make their data (transactions) immutable, so people will compete to get these block spaces, which is why we see very high gas prices on Ethereum L1.

02. Monolithic Blockchains

Monolithic Blockchains are blockchains that try to do consensus, execution, and data availability on L1 (main chain) at the same time. Basically, most blockchains to date, including the current PoW Ethereum chain, are monolithic blockchains.

The problem with monolithic blockchains is that they are bound by the "Blockchain Trilemma". Since the L1 layer of the same blockchain is responsible for providing the three underlying components that make a blockchain a "blockchain" (i.e. consensus, execution, and data availability), optimizing one property will impose constraints on the other two properties:

Want to have faster block times and larger block sizes to provide more block space? Then you can reduce the number of nodes that can keep up with the blockchain. This way, the slow computers in the world won't slow down the blockchain.

Want to confirm transactions faster? You can reduce the number of nodes, so that the total number of computers (nodes) that need to perform calculations is less. This way, we don't have a bunch of redundant computers doing the same computational work; but it also means that we can only trust a few nodes and believe that these few nodes doing the heavy computational work will not lie to the network.

Want to optimize security and decentralization? You can reduce the supply of block space, and lower the hardware requirements of nodes so that everyone can participate in the network as a node, but then your transactions will take longer to confirm.

Monolithic blockchains have gotten us so far, but now they are reaching the limits of scalability.

The era of monolithic blockchains is coming to an end.

The era of modular blockchains is coming. 03. Modular Blockchains Modular Blockchains take the three components of the current monolithic blockchain L1 (consensus, execution, and data availability) and divide them. Just like the division of labor, splitting these three components allows us to optimize each component and produce a better product, making the whole greater than the sum of its parts.

(1) Modular execution through Rollups

Rollups process transactions several orders of magnitude faster than the L1 main chain. By creating an off-chain transaction execution environment independent of Ethereum L1 and updating the state of L1 after processing transactions, Rollups do not have to bear the responsibility of consensus and data availability.

Rollups chains do not need to care about consensus and data availability like highly decentralized L1 chains; instead, Rollups chains are free to make any sacrifices in consensus and data availability because Rollups chains are cryptographically tied to Ethereum L1. In other words, Rollups are created by initiating a transaction on Ethereum L1 based on a set of rules, and Rollups promise to abide by this set of rules. From the moment a Rollup chain is initialized, it makes a cryptographic commitment to Ethereum L1 that it will abide by these rules.

The "initial commitment" of Rollups sets its own restrictions on how transactions are managed (i.e., Rollups promise to submit mathematical proof that all transactions are legal), and how the security of Ethereum L1 can be bridged to Rollups without porting L1's slow consensus and limited data availability to Rollups.

In this transaction that initializes the Rollup, any user is given the ability to withdraw their funds from the Rollup, which is called an "escape hatch", which means that when a Rollup network crashes or becomes malicious, the user can jump out of the escape hatch through an L1 transaction. A crashed Rollup network is like a broken elevator, except that the elevator has become a walking ladder.

Whether the Rollup is online and running normally, there is a "bridge" between the Rollup and Ethereum L1, which allows the settlement guarantee of Ethereum L1 to extend to Rollups.

The bridge connects the security and decentralization of Ethereum L1 to the transaction execution environment of Rollups.

With this bridge, each module of Ethereum complements the other; the security module (PoS) is added to the scalability module (Rollups). The properties of one module are integrated into the properties of the other, which is how Rollups achieve both scale and decentralization without compromise.

Rollups have almost zero maintenance costs, and the Rollups network only needs a small number of nodes to be active at any given time, and these nodes do not need to undertake expensive consensus mechanisms to ensure security. Ethereum L1 provides security and maintains decentralization, so Rollups do not need to do so.

Some types of Rollups can even have the same high performance as centralized servers. In fact, further innovations in Rollups will make them more performant than centralized databases.

(2) Modular security through PoS validators

The PoS consensus mechanism creates an intangible object responsible for providing security for the system, namely the virtual currency staked on the PoS network. This method of using the blockchain's native currency to verify the blockchain decouples the connection between physical (mining) hardware and network security.

Through the PoS mechanism, no specific computer is required to be responsible for network security. Now all computers can be used to be responsible for network security. Since ETH can be staked on any computer connected to the Internet, this reflects the value of providing security for the asset itself.

In PoS consensus, the physical (mining) capital cost of maintaining a PoW network is transformed into the cost of purchasing PoS tokens, thereby improving the capital efficiency of assets. Unlike physical mining hardware, PoS assets do not degrade over time, so PoS validators basically do not need to sell assets to cover operating costs.

In the case of the Ethereum PoS beacon chain, the economic cost of running a validator node is reduced to 32 ETH and a computer, which increases the total potential number of blockchain validators. While the cost of 32 ETH is still expensive (about $128,000 at the time of writing), it is an order of magnitude lower than the minimum viable PoW mining operation (tens of millions of dollars). In addition, decentralized staking pools like Lido and Rocketpool allow users to aggregate any amount of ETH for staking, making the 32 ETH limit no longer a hard limit.

The PoS network eliminates the hardware requirements for validating the network, making ordinary consumer-grade devices powerful enough to participate in the verification of the blockchain. This optimizes the connection between the blockchain network and the hardware. By minimizing the role of hardware devices, the accessibility of the blockchain can be maximized and the possibility for the maximum number of people to participate in blockchain verification is made possible. PoS minimizes the capital requirements for participating in network verification.

Since the launch of the Ethereum PoS beacon chain, Ethereum now has two homogeneous pools. When the new and old chains merge, the two pools become a modular network security pool, also known as the "validator pool", which will be the source of Ethereum's security.

Ethereum developers have expressed a desire to see 10 million ETH staked to be considered “secure”: 10 million ETH staked, 32 ETH per validator, which equates to 312,500 validators in the network.

Granulating Ethereum’s security into individual validator instances allows the beacon chain to direct those validator instances to wherever those resources are needed, giving Ethereum the greatest choice in how to allocate its security resources.

Having an available modular pool of security resources (i.e., validator pools) allows Ethereum to achieve modularity in data storage capabilities through sharding.

(3) Maximizing Data Availability: Sharding

Sharding will maximize the available block space in L1.

All blockchains have a supply of available security. Bitcoin’s security is the supply of SHA256 hashes that can be generated by the world (miners); PoS Ethereum’s security is the supply of Ethereum validators in the validator pool.

PoS Ethereum has a “validator pool” consisting of all validators, and all validators in the pool are randomly selected to validate Ethereum blocks. When more validators come online and provide their security to the Ethereum network (i.e., stake 32 ETH and promise to abide by the validation rules), this will make the Ethereum network more secure.

When shards are added to the network, it will also make Ethereum more scalable. Sharding allows for "security redistribution" across more shards, rather than concentrating the security of the entire system on one chain.

Having 30,000 validators (each staking 32 ETH) to protect a monolithic blockchain is an excess of security and an inefficient allocation of resources. By distributing these validators across multiple shards (64 shards will be launched initially), Ethereum L1 can create the scale of 64 Ethereum by distributing about 4,500 validators on each shard.

This makes Ethereum’s scalability positively correlated with its security, and while monolithic blockchains approach the limits of the “blockchain trilemma”, sharded blockchains invert the relationship between scale and security, turning their limiting factor into a growth factor.

Ethereum will initially implement 64 shard chains, but the goal is to increase to 1024 shard chains. As Moore’s Law continues, and all home computers become more powerful, the number and capacity of shard chains will increase.

Initially implementing 64 shard chains does not mean that we have increased Ethereum’s capacity by 64 times, but that we have 64 times more “Ethereum chains”, but each chain will be about 1/3 the size (capacity) of what it is now, so the overall size will increase by about 18 times, not 64 times.

However, as mentioned above, as physical hardware improves and the Ethereum validator pool grows, we will be able to increase the size and number of shards, tying Ethereum’s scalability to Moore’s Law.

The beacon chain is the “dispatcher” of system resources. Through random sampling, the beacon chain assigns each validator to a specific shard chain, where they will be responsible for verification. All validators are reshuffled every epoch (about 6.4 minutes).

04. Optimizing synergy between modules

The beauty of modular design is that optimization of each module amplifies the optimization of other modules. There are three synergies:

1. As more validators come online, modular PoS security is able to redistribute validators across more shard chains and is able to securely support more data. Better decentralization → greater scale.

2. More shard chains on L1 have a greater impact on the execution capacity of Rollups. Rollups can compress and add large amounts of data to L1 shard chains, so any additional space provided by shard chains will have a huge impact on the space available on Rollups. Greater scale → Faster execution.

3. The more net transaction activity that occurs on Rollups, the more total fees are paid when purchasing L1 block space. The total fees paid for block space increase the income of L1 validators. The more income validators have, the greater the incentive to attract more validators to join. Adding more validators on L1 adds more computational resources to the network, which can be used to create more shard chains. What does more shard chains mean? See #2 above.

More Scale, Faster Execution

By sharding Ethereum into 64 different data availability layers, we create more room for Rollups to deploy thousands of transaction batches. Sharding L1 will have a huge impact on the scalability of Rollups on L2. Since Rollups compress transactions into compact packets, any increase in data on L1 will create orders of magnitude more space on L2.

This is when Ethereum can achieve tiny transaction costs. Sharding Ethereum will open the floodgates for all Rollups: the increase in available blockspace can significantly reduce the fees of Rollups on shards.

Compressed Rollup transactions now have a much larger supply of available blockspace. Rollups will amortize L1 transaction costs across all of their users. For example, if it costs 1 ETH to deploy a large transaction bundle on Ethereum L1, the cost of 1 ETH will be amortized across the thousands of transactions in that bundle. By the time we have 64 shard chains to deploy transactions to, the cost per transaction should drop by orders of magnitude.

Once this is achieved, Rollups will be free to stop limiting the amount of blockspace available to them (as the current Arbitrum One network does), and let the engines of Rollups really start to work.

The combination of sharding and Rollups will allow computing resources to become an asset to the network, rather than a liability. More computers, regardless of their computing power, will always be able to contribute their resources to the network and make those resources effectively utilized, regardless of how much resources the computer can provide. A computer can be a Rollup validator and help compress and send the Rollup's data to L1, or it can contribute its resources to the L1 validator pool and help launch more shard chains.

In contrast, adding your node to a monolithic blockchain will create another bottleneck that the network must deal with. Understand that a monolithic blockchain cannot handle more transactions than a single node, and since all nodes of a monolithic blockchain need to process all transactions, adding your computer to a monolithic blockchain will only add another computer to the network that needs to be able to keep up with the entire network.

Monolithic Ethereum (left) vs. Modular Ethereum (right)

Economic Sustainability

Modular Ethereum is an economically sustainable Ethereum. This is the field of cryptoeconomics, and in addition to cryptography, we need to get the economics right.

In economics, Gresham's law is a monetary principle that states that "bad money drives out good money." When people are presented with two different currencies, they will store the one that has value and spend the one that has no value.

With fiat currencies, we see people moving to the currency that depreciates the least, which is the dollar, and in the cryptocurrency world, we ask "which cryptocurrency has appreciated the most in value?"

Bitcoiners are very excited about Bitcoin because it is the first cryptocurrency that promises to maintain its purchasing power through a hard cap on issuance (only 21 million BTC in total). As the economy around Bitcoin develops, Bitcoin is expected to become more scarce. The same BTC supply, but in a larger economy, equates to more scarce BTC.

Ethereans are excited about ETH and how much it is burned is proportional to the demand on the Ethereum network, and that ETH could become deflationary as more ETH is burned via EIP-1559 than is newly issued. A larger Ethereum economy means a higher rate of ETH being burned, making ETH increasingly scarce.

Transaction Fees = Monetary Premium

To put Gresham’s law into cryptoeconomic terms: the network needs to collect more transaction fees than it can pay to validators in the form of new coins.

Cryptoeconomic networks pay the network’s security providers (aka miners or validators) using transaction fees and new coins. The more transaction fees a blockchain network is able to collect, the fewer new coins it needs to issue.

In other words, the more transaction fees collected, the fewer coins need to be issued.

This is where the scalability problem of monolithic blockchains lies. Many blockchains promise low transaction fees and high throughput. But by doing so, these blockchains also promise to never create a meaningful fee market. This is because if you want block space to be cheaper, then you must not rely on transaction volume to pay for the security of the network; Therefore, you must rely on the issuance of new coins, which, to use the terminology of Gresham's Law, will make it a "bad money".

Below is an excerpt from an article written by Polynya:

Let’s compare Polygon PoS and Solana.

The Polygon PoS network collects about $50,000 in transaction fees per day, or $18 million per year. At the same time, Polygon is distributing more than $400 million in annual inflation rewards. This results in an incredible 95% net annual loss.

As for Solana, it collected only $10,000 in transaction fees per day for a long time, but as the speculative frenzy emerged, the network’s transaction fee income has increased significantly to $100,000 per day, or $36.5 million per year. But Solana is issuing an even more staggering $4 billion in annual inflationary rewards, resulting in a net loss of 99.2%.

Solana would need to reach 154,000 TPS at current transaction fees to break even, which is simply impossible given current hardware and bandwidth.

However, the bigger problem is that these additional transaction volume increases do not come without a cost - they add greater bandwidth requirements, greater state bloat, and generally higher system requirements.

The key feature of economic sustainability is that its effects compound in both directions.

A constrained L1 (such as the current Ethereum network) establishes a strong fee market. By limiting the amount of available block space, you can increase the decentralization of the blockchain (by reducing the hardware requirements of participating nodes) and fee revenue (by limiting the supply of available block space).

In the case of Ethereum, scarce block space generates higher fee revenue for the blockchain, which in turn leads to a higher ETH burn rate, making ETH more scarce and more valuable.

The more valuable a currency is, the fewer new coins need to be issued to obtain the same security. Therefore, when the value of the currency is high, the actual cost of security for the network is lower. In the paradigm of low security costs, the blockchain network will further reduce the net new issuance because you only need to issue fewer new coins. This will further increase the scarcity and value of the asset.

By contrast, blockchain networks that claim to offer a cheap fee environment are unable to collect any meaningful fee revenue. If the network cannot collect enough fee revenue, it must issue new coins to pay for security (i.e., to pay miners/validators who secure the network by issuing new coins). If the network ensures security by issuing new coins, then the currency will inflate and depreciate over time.

Inflation increases the supply of the currency over time and reduces its value. Depreciation means that the network must issue more new coins to pay for security costs. Further issuance will lead to further growth in supply and further depreciation of the currency, which is the beginning of a vicious inflationary cycle.

While speculation during bull markets can mask this effect for a while, the laws of economics cannot be avoided. A currency that issues new coins will not retain its value as well as a currency that is destroyed, and these two approaches will lead to very different futures.

There is a direct correlation between the throughput of L1 and the soundness of the network's native currency.

If you increase the throughput of a blockchain, you increase the inflation of the asset. Unfortunately, when you increase the throughput of a blockchain, you also reduce the ability of ordinary people to become validators.

This will divide the community around the blockchain into two categories of citizens: one group of citizens who have the ability to participate in validating the chain and are entitled to benefits; the other group of citizens who do not have this right and can only buy the coins sold to them by validators.

05. Putting it all together

Ethereum has a constrained L1, but with strong decentralization and efficient security. The limited block space of the current Ethereum network constrains L1, creating an expensive fee market, adding a monetary premium to ETH.

Sharding will increase the available L1 block space for Ethereum while increasing the security of the network. As Ethereum's validator pool expands (that is, the number of validators increases), the number of viable shard chains will also increase, making Ethereum a more scalable network while increasing its decentralization.

At the same time, Rollups create unconstrained execution environments that package and compress transactions into the smallest possible data packets. This unlocks new types of economic activity and allows a vibrant, cheap economy to flourish, which in turn increases the amount of net economic activity settled on L1. As more economic activity develops on Rollups, fees on Rollups will fall because these fees will be amortized across a wider range of participants. As more shard chains are added to Ethereum, and as the capacity of these shard chains becomes larger, fees on Rollups will fall further according to Moore's Law.

The beauty of modular design is that optimizations in each module amplify optimizations in the others:

· Increasing decentralization of the network through PoS increases the number of shard chains added to Ethereum;

· Adding more shard chains to Ethereum L1 can increase the scale of L2 Rollups by several orders of magnitude;

· The growth in the scale of Rollups will unlock new viable economic activities, which will ultimately increase the overall fees paid by Rollups to L1;

· Paying more overall fees to L1 will incentivize more people to become validators, making the validator pool of Ethereum L1 larger, allowing the network to create more shard chains.

And so on and so forth.

Monolithic blockchains that optimize for execution

With every bull run, there is a new wave of blockchains that choose to sacrifice decentralization to optimize the execution properties of the blockchain. These blockchains increase the block size and restrict more nodes to join, thereby providing a low-fee network for the booming bull market.

During bull markets, the Ethereum and Bitcoin networks become extremely congested because they optimize for decentralization, which provides an opportunity for monolithic blockchains that optimize for transaction execution.

But as mentioned above, monolithic blockchains that optimize for transaction execution have some drawbacks:They cannot collect fees in a meaningful way and sacrifice decentralization.

If this monolithic blockchain that optimizes execution transforms itself into an L2 network on a different L1 chain, then that chain can actually optimize execution better without having to deal with security and decentralization constraints. In this way, the network's native assets no longer need to be issued to pay for security because their integrity will now come from another L1 chain.

Removing inflation from the money supply schedule will have a huge impact on the long-term supply of their native assets.

Blockchains such as Solana, Binance Smart Chain (BSC), Avalanche, and Polygon may all need to transform themselves into Rollup chains to drive the long-term sustainability of their tokens. In fact, the sooner they transform themselves into Rollup chains, the scarcer their native assets will become.

Reasonable Conclusion

The cryptocurrency world is rife with tribalism and politics. One’s statements are often influenced by the tribe they belong to, and incentives and motivations are driven by pre-existing beliefs and biases.

Thankfully, code and math are not affected by these factors. We could rewrite the entire article above without using the word “Ethereum” as an unknown roadmap for a modular blockchain.

In fact, this architecture (i.e. modular architecture) is not unique to Ethereum. Rollups are not just an Ethereum thing. For example, Tezos is also embracing a Rollup-centric roadmap; NEAR is also designing for data availability sharding; Celestia is also building a security & data availability layer specifically for Rollups.

The point is, if we go back in time, or jump to different parallel universes, and roll the dice again 10,000 times, the cryptocurrency industry will find itself 99.9% of the time coming to the conclusion of modular design.

This is the most reasonable conclusion for the development of blockchain technology. The only reason this conclusion is "politically relevant" to Ethereum is that Ethereum is the only ecosystem that has provided enough funding for the relevant R&D work to enable us to achieve this goal so far.

Over time, we will see all L1s blockchains either evolve into modular design structures (i.e. limit L1 block space, push execution to Rollups, and increase the number of nodes) and compete in the world of global non-sovereign currencies, or they will abandon the burden of consensus and data and simply move their execution environment to another more decentralized blockchain (i.e. become another blockchain's L2 network).

The modular blockchain design also illustrates the necessity of decentralization as a key property of the blockchain to enable all other functions. Ethereum solves the scalability trilemma by increasing decentralization, rather than sacrificing it. Only by optimizing for decentralization can you gain the benefits of modularity described above.

If you embrace decentralization, then you can have anything.

Original link