An AWS researcher reveals go-verkle’s four major performance bottlenecks and has submitted optimization proposals

Gate News reports that on April 2, AWS’s core protocol researcher Ekleen Kaur and Everton Fraga presented at EthCC[9], conducting high-stress testing of go-verkle’s Verkle tree underlying implementation for Ethereum, revealing four major performance bottlenecks blocking the rollout of statelessness. Specifically: 1) During tree updates, a large number of shadow nodes are generated, causing state data redundancy to expand by about 35%; 2) The database index key design includes a large amount of duplicated path information, increasing storage overhead by about 50%; 3) When converting an in-memory tree structure for use in proofs, there are many memory copies rather than pointer references, causing deserialization speed to be 24 times slower than the optimized theoretical value, and under high load this may trigger proof verification timeouts; 4) The Verkle tree implementations in different-language clients such as Go, Rust, and Nimbus produce proofs that cannot be recognized by each other when handling specific edge cases, creating a risk of unexpected forks after an upgrade. The two researchers said AWS has submitted multiple optimization proposals to the Ethereum core repository, focusing on rebuilding deserialization logic to eliminate memory copies. After completion, they expect node execution efficiency to improve by more than 3 times, laying the foundation for the official integration of Verkle trees.