RadixKV - A Memory Efficient and High Performance Key-Value Store.

With the rapid development of the Internet of things (IoT), massive sensing data are continuously sent to data centers. Big sensing data call for efficient approaches on data storage and retrieval, because many IoT applications need real-time query performance. The Radix Tree is an efficient tree structure that has constant query time complexity which is much suitable for real-time query processing. However, it has high space cost and index building time, which will worsen the overall performance especially in update-intensive scenarios. In this paper, we propose a memory-efficient high-performance key-value system called RadixKV, which offers efficient improvements on the Radix Tree. We first analyze the online update performance of the Radix Tree and design an adaptive parallel index update strategy. Then, we propose Radix Array, which is a space optimized data structure for the Radix Tree. We conduct comparative experiments to evaluate the performance of RadixKV and compare RadixKV with previous works including Adaptive Radix Tree and Pre-order Radix Tree. The results show that RadixKV reduces 92.7% and 32.2% of space costs respectively, compared with the Adaptive Radix Tree and the Pre-order Radix Tree. In addition, RadixKV achieves the best write performance and has low index building time as well as query latency.