Ethereum’s next upgrade — EIP4844
Ethereum has recently been on an upgrade spree. The transition from proof-of-work to proof-of-stake went smoothly with the “Merge” in September 2022. “Shapella” in April 2023, allowed validators to withdraw staked ETH and staking rewards.
While Merge and Shapella were massive upgrades, neither of them directly solved Ethereum’s scaling problem which has been by far the biggest criticism of Ethereum. Executing only 15 transactions per second (tps) and having users pay exorbitant fees is hardly the description of a promising infrastructure for global finance. Layer 2s increase Ethereum’s capacity by ~100tps but that is clearly not enough.
While Ethereum still has the highest Total Value Locked (TVL) among smart-contract-supporting blockchains and is the preferred chain for DeFi, more transactions are processed by Solana. The reason for this becomes very obvious when one considers that the cost of processing 1mn NFTs in Ethereum is ~$33,000,000 compared to $133 in Solana.
EIP-4844, also called Proto-Danksharding, the next upgrade of Ethereum tackles scaling. It is expected to increase throughput to 1,000 tps and reduce fees by 10–100x times.
The ultimate scaling goal for Ethereum is Sharding. Full Sharding involves a complex build and therefore Ethereum is adopting a 3-step process to achieve the same:
Step 1 — Proto-Danksharding
Step 2 — Danksharding
Step 3 — Sharding
I’ll explain what each is in reverse.
SHARDING
Sharding means breaking the data into smaller units and dividing processing among different validators to increase throughput.
An easy way to understand sharding is to imagine a tolling station with 1 single booth. Every vehicle has to be cleared by that single booth and therefore the number of vehicles that can pass depends on the capacity of that single booth. Now imagine that 7 more toll booths are added. This immediately increases the speed to 8x as 8 different vehicles can be processed by 8 different booths simultaneously.
Applying the same concept, Ethereum initially planned to create 64 mini-chains instead of just one and have different validators process transactions on each mini-chains simultaneously, thus increasing capacity
DANKSHARDING
There are various complications to be addressed when executing Sharding and hence it was decided to implement danksharding first. Danksharding will increase Ethereum’s throughput to 100,000 tps.
The approach of Danksharding is very different. One way to scale a blockchain is to allow the processing of more transactions on the mainchain itself, i.e. on Layer 1. The other is to enable the processing of more transactions on Layer 2s i.e. Transactions can be processed on a faster Layer-2 chain and only transaction data is stored on the main chain. Danksharding supports the latter approach by reducing the cost of posting transaction data to the main chain.
To emphasize the difference between full sharding and danksharding- Full sharding will involve increasing the processing capacity of Layer 1 itself, while danksharding is an intermediate step where the processing capacity of Layer 2’s will increase.
Proto-danksharding and danksharding are key to the roll-up centric roadmap of Ethereum. The approach is to reduce complexity in the main chain and move computation to Layer 2s.
PROTO-DANKSHARDING
Proto-danksharding is a prototype of danksharding and will increase throughput to 1,000 tps.
It will reduce the cost of storing roll-up data on the main chain by introducing “Blob-Carrying Transactions”. Blob-carrying transactions are like regular transactions with the added functionality that they can carry BLOBs. BLOB stands for Binary Large Object and enables blocks to carry more data without incurring more costs.
Transactions will get executed on Layer 2s and the data will be included in Blobs that will be included in main chain blocks. Each blob is up to 125KB in size. On average a block will carry 8 blobs adding up to 1MB of data. This compares to an average block size today of 90KB.
The additional data can be carried without incurring additional costs because:
- Blob data does not touch the execution layer
Since the move to Proof of Stake, Ethereum has 2 clients — An Execution client i.e. Ethereum Virtual Machine (EVM), and a Consensus client i.e. Beacon Chain. Gas fees are incurred for processing power expended on EVM. Complex transactions use more processing power and are charged more gas fees. Beacon Chain earns fixed block rewards that are independent of the block size.
Blobs exist only on the Consensus Layer and not EVM and hence incur no execution fees making them far cheaper.
2. Blob data is periodically purged
Unlike calldata that is stored permanently on Layer 1, Blob data is deleted after some time. It has not been decided what this period will be (it could be a week or a month or 3 months or more) but there will be a time after which blob data will be removed from the blockchain reducing storage costs. The job of permanently storing data will become the responsibility of the roll-ups.
Danksharding superior to proto-danksharding
Danksharding will use the same formats and verification rules as proto-danksharding but will be superior in several ways.
The one key difference that deserves to be highlighted is that in proto-danksharding each node will need to carry the entire blob data, in danksharding, they may carry only a portion of the data.
This will be made possible by a combination of innovations like “Erasure Coding” and “Data Availability Sampling”. Essentially, data is first encoded and as long as at least 50% of the encoded data is made available, the entire data can be reconstructed. (It’s the same technology that makes it possible to recover data from scratched CDs).
This is important because 1MB of data per 12 seconds means that Ethereum will create 2.5TB of data annually. This is larger than the entire size of Ethereum today meaning that nodes will need to increase their storage capacity massively. Anyone familiar with the Bitcoin Wars knows that large-sized blocks are a recipe for centralization. This one property will therefore go a long way in protecting Ethereum from centralization.
This is the first of many steps to scaling
In addition to blob-carrying transactions, EIP 4844 also brings verification rules, multi-dimensional fee markets, and execution layer logic for full sharding. A lot of work, however, will remain for full sharding to be achieved including “Proof of Custody”, “ Proposer-Builder Separation”, “Data Availability Sampling”, “2D Sampling” etc
