An Adaptive Split Index Scheme for In-Memory Database Systems

With the rapid development of information technol- ogy, in-memory databases have significant performance advan- tages in large-scale, high-concurrency scenarios, but traditional index structures are prone to bottlenecks due to problems such as lock contention and transaction rollback. This paper proposes a dynamic contention detection (DCD) and adaptive split control (ASC) scheme based on Adaptive Radix Tree, which reduces lock contention and improves concurrency performance by monitor- ing node access contention and dynamically adjusting the index structure. Experiments show that in a high-concurrency, high- contention environment, DCDART has a throughput increase of about $\mathbf{1. 2}$ times compared to the original scheme. When pro- cessing high-concurrency writes and small-range query tasks, the optimized scheme shows significant performance improvements. This study provides a feasible index optimization idea for dealing with hot data and high-frequency concurrent transactions, and lays the foundation for subsequent applications in distributed environments.