Related Experiment Video
Updated: May 8, 2026

Microfluidic Imaging Flow Cytometry by Asymmetric-detection Time-stretch Optical Microscopy ATOM
Published on: June 28, 2017
High-Throughput and Memory-Efficient Pipeline Key-Value Store Architecture on FPGA
Xinshuo Wang1,2, Lei Liu1,2, Yifei Li1,2
1National Network New Media Engineering Research Center, Institute of Acoustics, Chinese Academy of Sciences, No. 21, North Fourth Ring Road, Haidian District, Beijing 100190, China.
None:
The increasing speed of network connections is placing increasing demands on the performance of network security and monitoring systems, where Key-Value Stores (KVSs) are becoming critical in network security applications. There is a compelling demand to enhance both the throughput and storage utilization of KVSs. The FPGA-based parallel architecture presents a remarkable opportunity to achieve outstanding performance and power efficiency. In this paper, we propose an FPGA-based implementation of KVSs using a multi-level multi-hash approach, which can effectively avoid false misses and false inserts, in addition to addressing skewed workloads. Decoupled storage exceeds 95% memory utilization, and the pipeline scheme achieves high performance, reaching 400 million requests per second (MRPS). The latency of insert, query, and delete operations is only 60 ns.

