Researchers at Anza and ETH Zurich have proposed a geographic scheduling system for Solana's validators designed to reduce block handover delays. The proposal, authored by Roger Wattenhofer and Quentin Kniep, would group nearby validators into consecutive block-production slots rather than using the current random ordering.
Under the existing system, validators are scheduled randomly to produce blocks during assigned windows. The new approach would maintain each validator's total number of production opportunities but reorder when those opportunities occur based on self-reported physical location. The aim is to reduce latency when one validator hands off its block to the next.
According to simulations using mainnet data from epoch 1038, the geographic grouping would cut mean handover delays between honest validators from 36.2 milliseconds to 17.0 milliseconds. The median delay across all handovers would fall from 23.4 milliseconds to 4.5 milliseconds. The researchers modeled these results using 661 validators and averaged results across five random seeds.
Unverifiable Location Data
A core limitation of the proposal is that it depends entirely on validator self-reporting of geographic coordinates. A companion specification would store these coordinates in validators' vote accounts with signed updates and geometric checks to confirm the reported point lies on Earth's surface. However, the actual physical location of validators' machines would remain unverifiable by the network.
The authors argue that misreporting location would typically harm a validator's own performance by placing it farther from the leaders it follows. Testing this economic incentive, they found that one validator in their sample could improve its mean handover delay by 2.7 milliseconds by falsely reporting a distant city, while other tested deceptions yielded negligible gains.
Security Trade-offs
The geographic grouping introduces security considerations around consecutive control of block production. Under the proposed three-window bin size, adjacent groups of validators can extend consecutive control beyond the configured bin size. In simulations, a 5% stake attacker positioned in Sydney could lead up to six consecutive windows by controlling two bins back to back.
The draft acknowledges that regional disruptions—whether from power outages, network issues, or jurisdictional action—could now affect multiple consecutive leaders, potentially causing longer skip sequences than the fully random schedule. The proposal also notes the possibility of more effective regional censorship during consecutive control periods.
Pending Implementation
Both the scheduling proposal and location-registration specification were introduced as pull requests on September 29 and remain open as of October 7. The scheduling proposal currently shows no reviews, while the location-registration pull request received conditional approval on October 5. The changes are proposed rules and modeled outcomes rather than results from a deployed system. The geographic schedule would require a feature gate and consensus activation to implement, with transition planned for two epochs after activation.
