Innovative architecture of single chip edge device based on virtualization technology

Abstract

Single-chip edge devices (SCEDs) are small in size and low in power consumption. They are especially suitable for smart nodes of Internet of Things (IoT) to form an edge sub-cloud (ESC). Unfortunately, SCED’s intensive computing ability is often unsatisfactory. Therefore, we need to migrate cloud computing virtualization technology into SCED, and manage its on-chip resources by cloud computing mode, so as to make its computing capability more powerful. The focus of this research is how to set up the on-chip cloud architecture of a domain general purpose integrated circuit (DGIC). The goal of this research is to provide a DGIC solution for SCED, which owns both of CPU customizable semantic processes and ASIC reconfigurable syntax elements. The cooperation of multiple SCED devices can form a strong edge sub-cloud ESC. Its on-chip cloud architecture can ensure the security and confidentiality of ESC. Based on the on-chip cloud structure of DGIC, we fabricated and tested a SCED sample in the acceleration domain of multi-island genetic algorithm. Its acceleration ratio was obviously better than the existing GPU + CPU approach, which has the advantages of small in size and low in power consumption.