An Ethereum prototype that divides blob-recovery duties among nodes has demonstrated significant efficiency gains in testing. Simulations across 1,000 nodes showed an 11–18× reduction in estimated reconstruction computing work, suggesting operators could reduce duplicated effort through a more modest change than previously considered.
Researcher Csaba Kiraly described the reduced design in a Sept. 3 report as a possible interim step toward a broader proposal known as RowDAS. The design assigns recovery duties without introducing new row-networking channels that the full RowDAS proposal would require.
How Ethereum Blobs Work
Blobs carry data used by layer-2 rollups. Ethereum's PeerDAS system allows nodes to download only partial blob data rather than complete copies. High-custody nodes hold at least 64 of the 128 data columns, providing enough information to rebuild missing blob data when needed. Supernodes store all 128 columns.
In the current system, many high-custody nodes perform the same reconstruction work independently. The reduced design instead assigns particular blobs to specific nodes first, allowing other nodes to receive the already-recovered data rather than repeating the same computational task.
Simulation Results
In one test configuration with four blobs and 10% supernodes, the estimated network-wide reconstruction cost fell from 48.6 CPU-seconds under the PeerDAS model to 2.75 CPU-seconds under the reduced design. With a 20% supernode share, the figures dropped from 91 to 6.6 CPU-seconds. These totals represent accumulated computing work across the network rather than recovery time.
Under the reduced variant, assigned nodes share recovered data through existing column-distribution channels. High-custody nodes retain a backup recovery role for any remaining missing data, preserving safeguards similar to PeerDAS.
Limitations and Next Steps
The measurements are limited to simulated networks using real cryptography. Kiraly reported no devnet results, and the full design's configuration with 128 row-subnets remains an extrapolation from smaller subnet counts. Larger simulations and real-network tests remain pending.
The full RowDAS proposal, specified in draft EIP-8371, would add another recovery route through row channels, allowing smaller nodes to pool data and reconstruct collectively. The reduced design retains dependence on high-custody nodes and cannot provide that additional resilience. The proposed split between duty assignment and row networking has yet to be incorporated into the draft's formal text.
