Ethereum research presented a state optimization experiment with remarkable results

Researcher of the weiihann execution layer shared the results of a large-scale experiment on optimizing the Ethereum state. As reported by Foresight, the study involved running mainnet load on the Geth client over the course of a year to compare traditional full nodes with a novel approach that stores only the active state, available for twelve months. The results of this experiment open new prospects for network scaling.

Revolutionary Reduction of Node Database Size

The most impressive outcome of the experiment is the radical decrease in hardware requirements. Comparing traditional full nodes with nodes that store only a one-year cycle of active states, researchers found a reduction in database size from 359 gigabytes to 81 gigabytes—almost a 77.5% decrease. The most notable results were observed with Trie storage optimization, demonstrating the effectiveness of a targeted approach to state data management.

This reduction has practical value: less disk space means lower entry barriers for running nodes, which could expand network decentralization and attract more participants.

Significant Improvements in Performance and Synchronization

The experiment revealed substantial temporal optimizations in node operation. Block re-execution time decreased by about 15%, and read operations showed the most significant improvement. Read latency at the P50 (median) level was reduced by 46%, while P99 (99th percentile) showed a 36% improvement. These metrics indicate acceleration in both normal and worst-case scenarios.

Another important indicator is tail latency optimization. Block insertion time at the P99 level decreased by 21%, which is critical for maintaining stable node synchronization under maximum network load.

Strategic Directions for Further Research

The team plans to expand the experiment by comparing results with other Ethereum clients, testing alternative state management cycles (such as six months), and exploring specialized contract storage cleanup strategies. Such comprehensive experiments will help determine optimal parameters for different types of nodes and network usage.