HIACS: A High-Performance Hybrid Index Structure Combining ART and Skip List for Non- Volatile Memory Storage Devices

Non-Volatile Memory (NVM) offers nanosecond latency and data persistence, enabling new index designs for large-scale key-value stores. However, its write asymmetry (high write latency, write amplification) and limited bandwidth challenge existing structures. The Adaptive Radix Tree (ART) excels in memory/NVM with prefix compression but suffers in NVM: its deeper tree than B+ trees causes long-key queries to traverse more NVM layers, degrading efficiency. Traditional Skip Lists shorten paths via hierarchy but fail to locate ART’s prefix branches due to random levels and full-key comparisons. This paper proposes HIACS, a hybrid NVM index integrating ART and Skip Lists. It uses a Prefix Hash Adjacency List (replacing Skip Lists) as H-Skip List to map key prefixes/suffixes to ART nodes for cross-hierarchical jumping and accurate branch localization; an Asynchronous Weak Consistency Strategy decouples leaf updates from SMOs with versioning to ensure eventual consistency and avoid blocking; and a Dynamic Monitoring Module tracks ART node states to optimize H-Skip List query parameters and reduce overhead. Experimental results under YCSB mixed workloads show HIACS outperforms mainstream NVM indexes (PACTree, LITS) by 23%–58% in throughput and over 40% in 99th percentile latency. It also has better space efficiency than B+ trees. HIACS addresses ART’s deep traversal and write amplification in NVM, providing a high-performance solution for large-scale data.