EVM Workloads in the Wild: Evidence for Multi-Dimensional Gas Metering, State Growth, Delayed Execution, and Parallelism

Jason Milionis; Kushal Babel; Lioba Heimbach

arxiv: 2606.19869 · v1 · pith:3N2W6S6Gnew · submitted 2026-06-18 · 💻 cs.DC

EVM Workloads in the Wild: Evidence for Multi-Dimensional Gas Metering, State Growth, Delayed Execution, and Parallelism

Lioba Heimbach , Kushal Babel , Jason Milionis This is my paper

Pith reviewed 2026-06-26 16:05 UTC · model grok-4.3

classification 💻 cs.DC

keywords EVM gas meteringstate growthblockchain workloadsstorage access patternsmulti-dimensional meteringparallel executionMEV transactionsDeFi gas sensitivity

0 comments

The pith

EVM gas estimates change across nearby historical states for 46% of Base transactions versus 13.9% on Ethereum.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper measures real transaction traces on Ethereum and Base to test the stability of execution conditions that single-scalar gas metering requires. It decomposes each transaction into opcode gas, intrinsic costs, refunds, and state deltas, then re-executes September 2025 transactions against older chain states to quantify how gas usage and storage patterns shift. The measurements show that resource mixes vary by chain and over time, persistent state accumulates hundreds of gigabytes, and gas estimates prove sensitive to state drift especially in MEV and DeFi workloads.

Core claim

The assumption that execution conditions remain stable enough to collapse costs into a single scalar with fixed prices does not hold: storage reads account for 29.2% and compute for 24.3% of execution gas on Base while storage writes dominate at 34.9% on Ethereum; Base exhibits 49.7% cold storage reads versus 39.6% on Ethereum; persistent state growth reaches 456 GB on Base and 38 GB on Ethereum; and gas estimates vary across nearby states for 46.0% of Base transactions compared with 13.9% on Ethereum.

What carries the argument

Trace-level decomposition of transactions into execution gas, intrinsic gas, refunds, and persistent state deltas, paired with cross-state re-execution to measure sensitivity of gas and storage access patterns.

If this is right

Multi-dimensional gas metering is required to reflect the varying shares of storage reads, writes, and compute.
Persistent state growth must receive explicit pricing because it is a permanent cost currently treated as transient.
Diverging storage access patterns across states limit the reliability of access lists.
State-dependent execution behavior complicates safe parallel execution of transactions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Workload predictors used by wallets and relayers may need state-aware adjustments to reduce failed or underpriced transactions.
L2 designs could benefit from chain-specific metering parameters given Base's higher observed sensitivity.
Models of delayed execution would need to incorporate state-drift variance when estimating final costs for users.
The data supports testing whether dynamic gas schedules that respond to recent state growth improve predictability.

Load-bearing premise

Re-executing September 2025 transactions on older historical states isolates the effect of state drift on gas usage and storage access patterns without artifacts from the execution environment or sampling bias.

What would settle it

A new trace study that re-executes a comparable sample of transactions and records gas estimates unchanged across nearby states for more than 85% of cases on both chains.

read the original abstract

Gas metering on EVM-compatible blockchains assumes that execution conditions are stable: that the resource mix is constant enough to justify collapsing execution costs into a single scalar with fixed relative prices, and that state drift between submission and execution does not materially alter a transaction's outcome. We measure the extent to which this assumption fails. We present a trace-level measurement study of EVM workloads on Ethereum (L1) and Base (L2) throughout 2025, sampling 3,000 blocks per day per chain. We decompose each transaction into opcode-level execution gas, intrinsic gas, refunds, and persistent state deltas. To measure state sensitivity, we re-execute transactions from September 2025 on older states and record how gas usage and storage access patterns change. We find the resource mix to be far from stable: on Base, storage reads and compute account for 29.2% and 24.3% of execution gas, while Ethereum devotes 34.9% to storage writes. Ethereum's gas limit doubling during 2025 shifted its own profile toward compute-heavier, Base-like patterns. Base also exhibits a higher fraction of cold storage reads (49.7% versus 39.6% on Ethereum). Persistent state growth, a permanent cost priced as a transient one, reaches 456 GB on Base versus 38 GB on Ethereum. Execution outcomes are equally unstable: gas estimates vary across nearby historical states for 46.0% of transactions on Base, compared to 13.9% on Ethereum, with especially high sensitivity for MEV and DeFi activity. Storage access patterns also diverge across states, limiting the effectiveness of access lists and complicating parallel execution. Our work provides an empirical foundation for multi-dimensional gas metering and explicit pricing of state growth. They show that state-sensitive execution behavior complicates workload estimation, directly affecting transaction predictability and user experience.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This paper delivers new 2025 numbers on state-driven gas variation (46% Base, 13.9% Ethereum) and state growth, but the re-execution method leaves room for artifacts.

read the letter

The key point is that the authors measured gas usage changes when re-executing September 2025 transactions against older chain states, finding variation in 46% of Base transactions versus 13.9% on Ethereum, plus state growth reaching 456 GB on Base. These are specific empirical figures not in the earlier literature they cite.

The work does a straightforward job of sampling 3,000 blocks daily per chain and breaking transactions down into opcode gas, refunds, and state deltas. The resource mix numbers (storage reads at 29.2% on Base, writes at 34.9% on Ethereum) and the cold-read fractions give a clearer picture of why single-scalar gas metering is shaky. The L1-L2 comparison and the observation that Ethereum's gas limit increase shifted its profile are useful context.

The main weakness is the re-execution step itself. Running later transactions on earlier states can pick up differences from client versions, gas schedule updates, or block context rules rather than pure state drift. The abstract supplies no validation of re-execution fidelity, error bars, or sampling checks, so the 46% figure rests on an assumption that may not hold cleanly. That concern is real but not fatal; it is the sort of thing referees can test.

This paper is for people working on EVM gas design, L2 scaling, or parallel execution. The data is concrete enough to matter for those discussions.

Send it to peer review. The measurements are new and the topic is practical; the method details can be tightened in revision.

Referee Report

2 major / 1 minor

Summary. The manuscript presents a trace-level empirical study of EVM workloads on Ethereum (L1) and Base (L2) throughout 2025. Sampling 3,000 blocks per day per chain, the authors decompose transactions into opcode-level execution gas, intrinsic gas, refunds, and persistent state deltas. They re-execute September 2025 transactions against older historical states to quantify state sensitivity, reporting that the resource mix is unstable (e.g., storage reads 29.2% and compute 24.3% of execution gas on Base), state growth reaches 456 GB on Base, and gas estimates vary across nearby states for 46.0% of Base transactions versus 13.9% on Ethereum, with higher sensitivity in MEV and DeFi activity. The work argues this evidence supports multi-dimensional gas metering and explicit state-growth pricing.

Significance. If the re-execution measurements are robust, the paper supplies concrete, large-scale observational data on resource-mix instability and state-drift effects that directly challenge the single-scalar gas model. The explicit percentages, cross-chain comparison, and identification of MEV/DeFi sensitivity provide a useful empirical foundation for proposals on access lists, parallel execution, and state pricing; the scale of the trace analysis is a strength.

major comments (2)

[Abstract] Abstract: the description of the sampling and re-execution approach supplies no details on error bars, exclusion criteria, or validation of re-execution fidelity (e.g., client version matching, gas-schedule invariance, or checks against original execution logs). This leaves the central claim that gas estimates vary for 46.0% of Base transactions only partially supported.
[state sensitivity measurement] The re-execution procedure for state sensitivity (September 2025 transactions on older states): the method implicitly assumes identical EVM client, gas schedule, and block-context rules at re-execution time. Any deviation in opcode semantics or post-2025 gas changes would confound protocol evolution with state drift and could inflate the reported 46.0% / 13.9% sensitivity figures.

minor comments (1)

[Abstract] Abstract, final sentence: 'They show that state-sensitive execution behavior...' uses an unclear pronoun; rephrase to 'The results show that...' or 'This study shows that...'.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on the abstract and re-execution methodology. We address each major comment below, indicating revisions where appropriate to strengthen the presentation of our empirical results.

read point-by-point responses

Referee: [Abstract] Abstract: the description of the sampling and re-execution approach supplies no details on error bars, exclusion criteria, or validation of re-execution fidelity (e.g., client version matching, gas-schedule invariance, or checks against original execution logs). This leaves the central claim that gas estimates vary for 46.0% of Base transactions only partially supported.

Authors: We agree the abstract is high-level and omits these specifics. The full manuscript details the sampling (3,000 blocks/day) and re-execution in the methods, including consistent client versions, gas-schedule matching to original execution, and cross-checks against execution logs. We will revise the abstract to briefly reference these validation steps and note that the measurements are deterministic traces (rendering statistical error bars inapplicable) while adding a short clause on exclusion criteria for invalid traces. This directly addresses the support for the 46.0%/13.9% figures. revision: yes
Referee: [state sensitivity measurement] The re-execution procedure for state sensitivity (September 2025 transactions on older states): the method implicitly assumes identical EVM client, gas schedule, and block-context rules at re-execution time. Any deviation in opcode semantics or post-2025 gas changes would confound protocol evolution with state drift and could inflate the reported 46.0% / 13.9% sensitivity figures.

Authors: This is a valid concern. Our procedure used the identical EVM client binary and gas schedule from the September 2025 execution window for all re-executions to isolate state effects; block-context rules were preserved via the historical state snapshots. We will add an explicit methods paragraph confirming the absence of gas-schedule or opcode changes in the study period and describing the client-version matching process. This clarification will prevent misinterpretation of the sensitivity results as confounded by protocol evolution. revision: yes

Circularity Check

0 steps flagged

No circularity: purely empirical trace measurements with no derivations or fitted parameters

full rationale

The paper conducts a measurement study by sampling blockchain traces and re-executing transactions against historical states. All reported figures (e.g., 46.0% gas variation on Base, resource mix percentages, state growth in GB) are direct observations from public data. No equations, ansatzes, uniqueness theorems, or self-citations appear in the provided text; the central claims do not reduce to any fitted input or self-referential definition. The work is self-contained against external blockchain traces.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on domain assumptions about sampling representativeness and re-execution fidelity; no free parameters or invented entities are introduced.

axioms (1)

domain assumption Sampling 3,000 blocks per day per chain yields a representative view of workloads throughout 2025.
Basis for all reported percentages and growth figures.

pith-pipeline@v0.9.1-grok · 5896 in / 1136 out tokens · 25439 ms · 2026-06-26T16:05:26.297324+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

15 extracted references · 5 canonical work pages

[1]

The opbench ethereum opcode benchmark framework: Design, implementation, validation and experiments

2 Amjad Aldweesh, Maher Alharby, Maryam Mehrnezhad, and Aad van Moorsel. The opbench ethereum opcode benchmark framework: Design, implementation, validation and experiments. Performance Evaluation, 146:102168, 2021.doi:10.1016/j.peva.2020.102168. 3 Guillermo Angeris, Theo Diamandis, and Ciamac Moallemi. Multidimensional blockchain fees are (essentially) o...

work page doi:10.1016/j.peva.2020.102168 2021
[2]

URL:https://eips.ethereum.org/EIPS/ eip-150

Created: 2016-09-24. URL:https://eips.ethereum.org/EIPS/ eip-150. 8 Vitalik Buterin. A state expiry and statelessness roadmap. Ethereum Foundation notes,

2016
[3]

URL:https://vitalik.eth.limo/general/2024/10/26/futures5.html

Blog post. URL:https://vitalik.eth.limo/general/2024/10/26/futures5.html. 10 Vitalik Buterin, Eric Conner, Rick Dudley, Matthew Slipper, Ian Norden, and Abdelhamid Bakhta. EIP-1559: Fee market change for ETH 1.0 chain. Ethereum Improvement Proposals, April

2024
[4]

URL:https://eips.ethereum.org/EIPS/eip-1559

Created: 2019-04-13. URL:https://eips.ethereum.org/EIPS/eip-1559. 22 EVM Workloads in the Wild 11 Vitalik Buterin and Martin Swende. EIP-2929: Gas cost increases for state access opcodes. Ethereum Improvement Proposals, September

2019
[5]

URL: https: //eips.ethereum.org/EIPS/eip-2929

Created: 2020-09-01. URL: https: //eips.ethereum.org/EIPS/eip-2929. 12 Vitalik Buterin and Martin Swende. EIP-2930: Optional access lists. Ethereum Improvement Proposals, August

2020
[6]

URL: https://eips.ethereum.org/EIPS/ eip-2930

Created: 2020-08-29. URL: https://eips.ethereum.org/EIPS/ eip-2930. 13 Vitalik Buterin and Martin Swende. EIP-3529: Reduction in refunds. Ethereum Improve- ment Proposals, April

2020
[7]

URL:https://eips.ethereum.org/EIPS/ eip-3529

Created: 2021-04-22. URL:https://eips.ethereum.org/EIPS/ eip-3529. 14 Danilo Rafael de Lima Cabral, Pedro Antonino, and Augusto Sampaio. Demystification and near-perfect estimation of minimum gas limit and gas used for ethereum smart contracts. Journal of Cloud Computing, 2025.doi:10.1186/s13677-025-00751-y. 15 Category Labs. Monad initial spec proposal. ...

work page doi:10.1186/s13677-025-00751-y 2021
[8]

Designing multi- dimensional blockchain fee markets

18 Theo Diamandis, Alex Evans, Tarun Chitra, and Guillermo Angeris. Designing multi- dimensional blockchain fee markets. In5th Conference on Advances in Financial Tech- nologies (AFT 2023), Leibniz International Proceedings in Informatics (LIPIcs),

2023
[9]

19 Dune Analytics

doi:10.4230/LIPIcs.AFT.2023.4. 19 Dune Analytics. Spellbook: A community maintained library of blockchain analytics.https: //github.com/duneanalytics/spellbook,

work page doi:10.4230/lipics.aft.2023.4 2023
[10]

Atomic mev table.https://dune.com/queries/3493305,

25hildobby. Atomic mev table.https://dune.com/queries/3493305,

work page arXiv
[11]

2020.24267

doi:10.14722/ndss. 2020.24267. L. Heimbach, K. Babel, J. Milionis 23 31 Yanjing Ren, Jia Zhao, Jingwei Li, and Patrick P. C. Lee. An analysis of ethereum workloads from a key-value storage perspective. In2025 IEEE International Symposium on Workload Characterization (IISWC), pages 394–406, Irvine, CA, USA, October 2025.doi:10.1109/ IISWC66894.2025.00040. ...

work page doi:10.14722/ndss 2020
[12]

URL:https://eips.ethereum

Created: 2019-03-28. URL:https://eips.ethereum. org/EIPS/eip-1884. 38 Wei Tang. EIP-2200: Structured definitions for net gas metering. Ethereum Improvement Pro- posals, July

2019
[13]

URL:https://eips.ethereum.org/EIPS/eip-2200

Created: 2019-07-18. URL:https://eips.ethereum.org/EIPS/eip-2200. 39 Toni Wahrstätter. Execution dependencies, April

2019
[14]

URL: https://ethresear.ch/t/execution-dependencies/22150

Ethereum Research forum post. URL: https://ethresear.ch/t/execution-dependencies/22150. 40 Fei Wu, Danning Sui, Thomas Thiery, and Mallesh Pai. Measuring CEX-DEX Extracted Value and Searcher Profitability: The Darkest of the MEV Dark Forest. In7th Conference on Advances in Financial Technologies (AFT 2025),

2025
[15]

24 EVM Workloads in the Wild A Daily State Growth Distribution Table 8 reports the distribution of daily storage growth in 2025, broken down into storage slots, contract bytecode, EOA state, and contract account state. For both chains, daily storage growth varies substantially, highlighting the mismatch between short-term per-block gas limits and the long...

2025

[1] [1]

The opbench ethereum opcode benchmark framework: Design, implementation, validation and experiments

2 Amjad Aldweesh, Maher Alharby, Maryam Mehrnezhad, and Aad van Moorsel. The opbench ethereum opcode benchmark framework: Design, implementation, validation and experiments. Performance Evaluation, 146:102168, 2021.doi:10.1016/j.peva.2020.102168. 3 Guillermo Angeris, Theo Diamandis, and Ciamac Moallemi. Multidimensional blockchain fees are (essentially) o...

work page doi:10.1016/j.peva.2020.102168 2021

[2] [2]

URL:https://eips.ethereum.org/EIPS/ eip-150

Created: 2016-09-24. URL:https://eips.ethereum.org/EIPS/ eip-150. 8 Vitalik Buterin. A state expiry and statelessness roadmap. Ethereum Foundation notes,

2016

[3] [3]

URL:https://vitalik.eth.limo/general/2024/10/26/futures5.html

Blog post. URL:https://vitalik.eth.limo/general/2024/10/26/futures5.html. 10 Vitalik Buterin, Eric Conner, Rick Dudley, Matthew Slipper, Ian Norden, and Abdelhamid Bakhta. EIP-1559: Fee market change for ETH 1.0 chain. Ethereum Improvement Proposals, April

2024

[4] [4]

URL:https://eips.ethereum.org/EIPS/eip-1559

Created: 2019-04-13. URL:https://eips.ethereum.org/EIPS/eip-1559. 22 EVM Workloads in the Wild 11 Vitalik Buterin and Martin Swende. EIP-2929: Gas cost increases for state access opcodes. Ethereum Improvement Proposals, September

2019

[5] [5]

URL: https: //eips.ethereum.org/EIPS/eip-2929

Created: 2020-09-01. URL: https: //eips.ethereum.org/EIPS/eip-2929. 12 Vitalik Buterin and Martin Swende. EIP-2930: Optional access lists. Ethereum Improvement Proposals, August

2020

[6] [6]

URL: https://eips.ethereum.org/EIPS/ eip-2930

Created: 2020-08-29. URL: https://eips.ethereum.org/EIPS/ eip-2930. 13 Vitalik Buterin and Martin Swende. EIP-3529: Reduction in refunds. Ethereum Improve- ment Proposals, April

2020

[7] [7]

URL:https://eips.ethereum.org/EIPS/ eip-3529

Created: 2021-04-22. URL:https://eips.ethereum.org/EIPS/ eip-3529. 14 Danilo Rafael de Lima Cabral, Pedro Antonino, and Augusto Sampaio. Demystification and near-perfect estimation of minimum gas limit and gas used for ethereum smart contracts. Journal of Cloud Computing, 2025.doi:10.1186/s13677-025-00751-y. 15 Category Labs. Monad initial spec proposal. ...

work page doi:10.1186/s13677-025-00751-y 2021

[8] [8]

Designing multi- dimensional blockchain fee markets

18 Theo Diamandis, Alex Evans, Tarun Chitra, and Guillermo Angeris. Designing multi- dimensional blockchain fee markets. In5th Conference on Advances in Financial Tech- nologies (AFT 2023), Leibniz International Proceedings in Informatics (LIPIcs),

2023

[9] [9]

19 Dune Analytics

doi:10.4230/LIPIcs.AFT.2023.4. 19 Dune Analytics. Spellbook: A community maintained library of blockchain analytics.https: //github.com/duneanalytics/spellbook,

work page doi:10.4230/lipics.aft.2023.4 2023

[10] [10]

Atomic mev table.https://dune.com/queries/3493305,

25hildobby. Atomic mev table.https://dune.com/queries/3493305,

work page arXiv

[11] [11]

2020.24267

doi:10.14722/ndss. 2020.24267. L. Heimbach, K. Babel, J. Milionis 23 31 Yanjing Ren, Jia Zhao, Jingwei Li, and Patrick P. C. Lee. An analysis of ethereum workloads from a key-value storage perspective. In2025 IEEE International Symposium on Workload Characterization (IISWC), pages 394–406, Irvine, CA, USA, October 2025.doi:10.1109/ IISWC66894.2025.00040. ...

work page doi:10.14722/ndss 2020

[12] [12]

URL:https://eips.ethereum

Created: 2019-03-28. URL:https://eips.ethereum. org/EIPS/eip-1884. 38 Wei Tang. EIP-2200: Structured definitions for net gas metering. Ethereum Improvement Pro- posals, July

2019

[13] [13]

URL:https://eips.ethereum.org/EIPS/eip-2200

Created: 2019-07-18. URL:https://eips.ethereum.org/EIPS/eip-2200. 39 Toni Wahrstätter. Execution dependencies, April

2019

[14] [14]

URL: https://ethresear.ch/t/execution-dependencies/22150

Ethereum Research forum post. URL: https://ethresear.ch/t/execution-dependencies/22150. 40 Fei Wu, Danning Sui, Thomas Thiery, and Mallesh Pai. Measuring CEX-DEX Extracted Value and Searcher Profitability: The Darkest of the MEV Dark Forest. In7th Conference on Advances in Financial Technologies (AFT 2025),

2025

[15] [15]

24 EVM Workloads in the Wild A Daily State Growth Distribution Table 8 reports the distribution of daily storage growth in 2025, broken down into storage slots, contract bytecode, EOA state, and contract account state. For both chains, daily storage growth varies substantially, highlighting the mismatch between short-term per-block gas limits and the long...

2025