Performance Analysis of Adaptive Resource Allocation Scheme for OpenCL-based FPGA Virtualization System

Abstract

A recent advance in the performance of FPGA devices is creating an industry trend toward an energy efficient, high-performance computing paradigm. In response to this trend, major cloud providers have begun to offer FPGA-as-a-Service (FaaS). Unfortunately, most of the current FPGA services and researches are inefficient in resource utilization, that they assign an FPGA per process which cannot be shared among users. In this paper, we propose a partial reconfiguration (PR) based virtualized FPGA (vFPGA) system with OpenCL compatibility. Moreover, we propose a run-time management system and dynamic resource allocation scheme to efficiently allocate resources in the system. With our system, FPGA resources can be fully utilized and shared by multiple users. Experimental results show that our proposed vFPGA system can accomodate multiple users with high resource utilization and low service latency. We suggest that our system is suitable for the current trend of cloud computing towards the mobile edge computing (MEC) cloud.