HART: A Concurrent Hash-Assisted Radix Tree for DRAM-PM Hybrid Memory Systems

Persistent memory (PM) exhibits a huge potential to provide applications with a hybrid memory system where both DRAM and PM are directly connected to a CPU. In such a system, an efficient indexing data structure such as a persistent tree becomes an indispensable component. Designing a capable persistent tree, however, is challenging as it has to ensure consistency, persistence, and scalability without substantially degrading performance. Besides, it needs to prevent persistent memory leaks. While hash table has been widely used for main memory indexing due to its superior performance in random query, ART (Adaptive Radix Tree) is inherently better than B/B^+-tree in most basic operations on both DRAM and PM. To exploit their complementary merits, in this paper we propose a novel concurrent and persistent tree called HART (Hash-assisted ART), which employs a hash table to manage ARTs. HART employs a selective consistency/persistence mechanism and an enhanced persistent memory allocator, which can not only optimize its performance but also prevent persistent memory leaks. Experimental results show that in most cases HART significantly outperforms WOART and FPTree, two state-of-the-art persistent trees. Also, it scales well in concurrent scenarios.