AWS Researchers Identify Performance Bottlenecks in Ethereum's Verkle Tree Implementation

Amazon's cloud computing division, AWS, has conducted high-pressure testing on the go-verkle implementation of Ethereum's Verkle tree at the EthCC[9] conference. According to Foresight News, researchers Ekleen Kaur and Everton Fraga revealed four major performance bottlenecks hindering the stateless implementation. Firstly, during tree updates, numerous shadow nodes are created that do not exist logically but occupy physical space, leading to approximately 35% redundancy in state data. Secondly, the design of database index keys includes a significant amount of duplicate path information, increasing storage overhead by about 50%. Thirdly, converting the memory tree structure to proofs involves extensive memory copying rather than pointer referencing, resulting in deserialization speeds that are 24 times slower than optimized theoretical values, potentially causing proof verification timeouts under high load. Lastly, the Verkle tree implementations in different language clients such as Go, Rust, and Nimbus generate proofs that are not mutually recognizable in specific edge cases, posing a risk of unexpected forks after upgrades. The researchers noted that AWS has submitted several optimization proposals to the Ethereum core repository, focusing on restructuring deserialization logic to eliminate memory copying. Once completed, node operational efficiency is expected to improve by over three times, laying the groundwork for the formal integration of Verkle trees.