학술논문

Data Matching Engine for Efficient Big Data Searching Systems Using a RRAM Based Novel Computing Paradigm
Document Type
Conference
Source
2023 Silicon Nanoelectronics Workshop (SNW) Silicon Nanoelectronics Workshop (SNW), 2023. :5-6 Jun, 2023
Subject
Components, Circuits, Devices and Systems
Computing and Processing
Engineered Materials, Dielectrics and Plasmas
Resistance
Energy consumption
Voltage measurement
Array signal processing
Switches
Big Data
Silicon
Language
ISSN
2161-4644
Abstract
For the first time, we propose and demonstrate an efficient data matching engine using a novel computing paradigm based on resistance switching computing in RRAM crossbar array. In the engine, in-memory logic operations, such as NAND, XOR and multi-input AND, are in-situ executed by RRAM crossbar array to perform the matching operation. Leveraging the reduced data movement and computing steps, together with parallel computing, RRAM-based data searching system achieves 1600 times less runtime and 6.1 times better energy consumption performance than a GPU-based solution.