Ira: Efficient Transaction Replay for Ethereum (Artifact Data)

Data artifact accompanying the USENIX Security 2026 paper *"Ira: Efficient Transaction Replay for Ethereum"* by Adithya Bhat, Harshal Shah, and Mohsen Minaei (preprint: arXiv:2601.21286). Source code, analysis scripts, benchmark CSVs, and the end-to-end reproduction README live at the companion GitHub repository: https://github.com/adithyabhatkajake/ira-rs (tagged release `v1.0-usenixsec26`, commit `3348194`). This Zenodo record provides a DOI-bearing archive of the three large data assets that are too big to ship in the source repository, satisfying the USENIX "Artifact Available" requirement. ## Contents - **`ira-hints.tar`** (14 GB). SHA-256: `3b3d0dc9a655091c49d2c616b3db8034c724d866fecdd99db89b756b83860b54`. Hint database emitted by `reth-primary` over Ethereum mainnet blocks 24,019,447–24,120,246 (100,800 blocks, Dec 15–29, 2025). Contains both the consolidated `hints.redb` (~8 GB, consumed by `reth-backup`) and the per-block `batch_*/<block>.hint.zst` files (~6 GB, consumed by the source-byte-distribution analysis script). Both formats are load-bearing. - **`ira-parquet-traces.tar`** (6.8 GB). SHA-256: `2e4a663aebc78e839179a360265ebe1842444ac5b830516b67a12b5a7794873d`. Eleven Apache Parquet shards (Zstd-compressed) emitted by `ira-trace-collector`. Per-row schema `(block_number, tx_index, op_type, address, slot, value, source_byte)`. Consumed by the §3 Ethereum case-study scripts under `scripts/analyze-*.py` and `scripts/generate-*.py`. - **`ira-state-hashes.tar`** (295 MB). SHA-256: `aa8456456414ef1963390b7b322d29453433fbc4bfd4f7773667e890b434d943`. Per-block deterministic post-state fingerprints emitted by `reth-primary`. Used by `reth-backup --state-hash-dir` to verify post-block state-root correctness. - **`ira-source.tar`** (142 MB). SHA-256: `8c1f1e6d8010db49d56311bb5e938b15ce6c1bfd4f35f875af0e7b20fbf295d7`. Complete source archive of the `v1.0-usenixsec26` tagged release: the Rust workspace (`reth-primary`, `reth-baseline`, `reth-backup`, `ira-trace-collector`, `io-trace`, `block-size`), Python analysis and plotting scripts (`scripts/`), paper-side reproduction harness (`artifact/`), the 46 evaluation CSVs (`data/`), `README.md`, `LICENSE` (MIT), `AUTHORS`, and `requirements.txt`. Built via `git archive v1.0-usenixsec26`; no build artifacts, no `.git/`, no uncommitted state. - **`SHA256SUMS`** — manifest of the three hashes above. Verify with `shasum -a 256 -c SHA256SUMS`. ## What is *not* deposited The full reth archive datadir (~3 TB) is **not** redistributed here. Anyone can reproduce it by running reth at the pinned commit against mainnet up to block 24,120,246; see the GitHub README for the exact command. ## Reproduction See `README.md` in the GitHub repository for the end-to-end build / run / regenerate-figures recipe. The CSVs bundled in the GitHub repository already regenerate every paper macro and figure without this Zenodo data; the data here is only required for the full re-execution path that starts from a synced reth archive node.