EIP-4844: Ethereum Proto-Danksharding

Updated 12 October 2025

EIP-4844 is a protocol upgrade enabling transient, large blob data storage that decouples on-chain execution from data availability for scalable rollup processing.
It introduces a dual fee market with separate blob gas pricing, dynamically adjusted via exponential rules to optimize transaction costs and resource allocation.
Empirical findings show reduced data posting costs and improved rollup efficiency, while also highlighting challenges in optimal block packing and fee market performance.

Ethereum Improvement Proposal 4844 (EIP-4844), widely known as "Proto-Danksharding," is a protocol upgrade implemented on the Ethereum network to address the scalability and cost-efficiency constraints imposed by on-chain data availability. By introducing a new transaction type incorporating large, transient "blobs" of off-chain data, EIP-4844 unleashes the potential for high-throughput rollup-centric computation, reduced gas fees, and sets the stage for future full data sharding. The following sections elaborate on the design, economic mechanisms, empirical findings, and systemic impacts of EIP-4844, as documented in the corresponding technical literature.

1. Design Principles and Transaction Mechanism

EIP-4844 introduces "blob-carrying transactions" (BTXs, often "type-3 transactions"), which allow the inclusion of up to six blobs (each up to 128 KB) per transaction. Unlike calldata, blobs are not stored permanently on-chain, but are referenced and verified via commitment schemes such as KZG, with the actual data held off-chain for a retention window of roughly 18 days (Soltani et al., 17 Sep 2024, Heimbach et al., 18 Feb 2025).

The dual-market design leverages a separate "blob gas" resource, priced independently from EVM computation/stored gas (see Table below). The traditional EIP-1559 gas mechanism for execution persists, while the "blob gas base fee" is dynamically adjusted via an exponential update rule to target three blobs per block:

Resource	Pricing Mechanism	Storage Duration
Standard Gas	EIP-1559 base+priority fee	Permanent
Blob Gas (4844)	Base fee: exponential adjustment	~18 days (transient)

Blob fees are updated as:

$\text{blob\_base\_fee}(n) = \min\_fee \cdot \exp\left( \frac{\text{total\_excess\_gas}(n-1)}{\text{update\_fraction}} \right)$

where excess gas tracks deviation from the protocol's blob-per-block target (Heimbach et al., 18 Feb 2025).

2. Economic Models for Rollup Strategies

The principal intended beneficiary of EIP-4844 is the rapidly developing rollup ecosystem, which depends critically on cost-effective data availability (DA). From a rollup's perspective, posting data via blobs involves a trade-off between a fixed posting cost and delay cost. If transaction arrival rates are low, filling a blob (to amortize the fixed posting fee) can cause unacceptable delay, whereas small rollups posting partial blobs waste space and pay disproportionately higher fees (Crapis et al., 2023, Lee, 5 Oct 2024). The per-transaction cost under blob posting is given by:

$\text{Tr}_B(t) = \frac{P_0 G + B}{Rt} + \frac{a t}{2}$

where $R$ is the transaction rate, $B$ is the blob cost, $t$ is the posting interval, and $a$ quantifies delay cost.

The optimal strategy balances delay and posting costs:

$t^* = \sqrt{\frac{2(P_0G + B)}{aR}}$

When $R$ is low, many rollups prefer to use traditional calldata ("main market") for data posting. The equilibrium blob price, derived for a market with $k$ blobs per time unit and rollup rates $R_i$ , is:

$B = \frac{a\left(\sum_i \sqrt{R_i}\right)^2}{2k^2}$

3. Fee Market Dynamics and Optimization

The empirical literature highlights frailties in the blob fee market design during real network stress (Heimbach et al., 18 Feb 2025, Huang et al., 6 Nov 2024). Specifically, blocks may not be optimally "packed" with blob transactions, resulting in measurable builder fee loss (up to 70% fee loss in congestion events). This is due to the combinatorial nature of block inclusion—each block can include at most six blobs, and transaction inclusion resembles a multi-dimensional knapsack problem.

A central inefficiency arises from "subset bidding," where current transaction formats compel all-or-nothing blob inclusion. Thus, transactions requesting more blobs than can be accommodated may be excluded, even if partial inclusion would maximize block builder revenue. Proposed improvements include support for permutation bidding or transaction splitting to enable optimal block packing.

Furthermore, the interdependence between builder and rollup strategies is pronounced: when builders maximize inclusion of high-value (e.g., mempool) transactions, rollups must overbid for inclusion; if rollups optimize costs, builder profit falls (Huang et al., 6 Nov 2024). The empirical analysis demonstrates that nearly 30% of builder blocks and over 70% of rollup type-3 transactions are economically inefficient.

4. Delay Analysis and System Throughput

EIP-4844's blob-carrying transaction queue can be rigorously modeled as an M/D $^B$ /1 queue (Soltani et al., 17 Sep 2024). Here, $B$ is the blob capacity per block, arrivals are Poisson, and service time per block is deterministic. Key findings from this model:

At low load, BTX delay approaches half a block interval (service time).
Under load $\rho = \lambda \tau / B$ approaching 1, delay grows rapidly, especially if BTXs carry multiple blobs.
Higher-frequency, smaller blob transactions minimize system delay; infrequent, larger blob transactions increase delay due to reduced effective batch size.

Projections for future upgrades suggest that delay will become a central parameter balancing system scalability, gas cost, and user experience as demand increases.

The fixed 128 KB blob size generates operational challenges for small rollups with low data throughput. Blob sharing—multiple rollups aggregating into a single blob—directly addresses the cost inefficiency and DA latency dilemma (Lee, 5 Oct 2024, Crapis et al., 2023). Simulation and empirical studies demonstrate that:

Blob sharing can yield >85% cost reduction for small rollups.
DA service quality, as measured by the frequency and regularity of data posting, is significantly improved.
The smoothing effect on the blob base fee reduces exposure to exponential fee increases (triggered beyond three blobs per block).

Optimally, blob sharing is governed by a Nash bargaining solution, where joint participants split savings in a manner that is Pareto efficient and satisfies standard fairness axioms.

6. Data Availability and Decentralization in Layer 2

EIP-4844, by enabling cost-effective off-chain data publication, raises critical challenges for maintaining data availability and L2 decentralization (Huang et al., 16 Mar 2024). To counter "lazy validator" and data withholding risks in L2, contemporary research has introduced:

Proof of Download: Batches include hidden states computed from previous transactions, compelling nodes to retrieve historical data.
Proof of Existence: KZG polynomial commitments and cryptographic challenge-response protocols enforce the persistent on-demand availability of old data.
Role Separation: Splitting the L2 node population into low-resource proposers and computation-intensive builders enhances decentralization without compromising performance.
Proof of Luck: Random assignment mechanisms for builder role selection reduce the risk of proposer-builder collusion and MEV extraction.

These measures are crucial as the volume of off-chain rollup data increases with greater blob utility, preserving both scalability and system security.

7. Impact on Consensus, System Security, and Future Sharding

Empirical evaluations post-EIP-4844 implementation show:

A doubling of fork rate in the consensus layer (from 3.097 to 6.707 per 2000 slots), corresponding to a 140 ms average increase in block synchronization time (Park et al., 6 May 2024). Receive time for slot data increased by 81 ms, attributable to blob processing overhead.
An average reduction in data posting costs for rollups (1.304 to 0.231 ETH per MiB; 82% decrease), with mean data per block increasing from 0.084 to 0.183 MiB.
Marked volatility in blob gas fee markets, although the mechanism is more responsive to short-term demand than the traditional fee market.

These findings highlight both the realized and hypothetical trade-offs between scalability, consensus robustness, and user cost. Proto-Danksharding, as realized in EIP-4844, establishes a framework for future advanced sharding solutions, with the potential for further optimization in multidimensional fee markets and block inclusion algorithms.

In summary, EIP-4844 marks a pivotal transition in Ethereum's technical trajectory. By decoupling data availability from on-chain execution, and by operationalizing transitory, low-cost blobs with an independent fee market, it substantively improves the economic viability of rollup-based scaling. This results in complex resource allocation and strategic interdependence in the builder-rollup ecosystem, introduces new optimization challenges, and motivates advances in data availability schemes and decentralized validator participation. Empirical findings post-deployment reveal both substantial gains in efficiency and persistent inefficiencies requiring further refinement, particularly regarding fee market design and block packing. As such, EIP-4844 not only solves current scalability bottlenecks but also establishes the foundational mechanisms for the next generation of sharded and rollup-centric blockchain protocols.