NoC based virtualized accelerators for cloud computing

Abstract : Hardware accelerators (HwAcc) provide good performance in computation intensive applications. Integrating hardware accelerators in a cloud environment is the optimal way to improve the quality of service. However, mapping all possible application statically into the reconfigurable fabric of the FPGA is rather impractical and prohibitively expensive in terms of resource and power consumption. This problem can be alleviated via time multiplexing the access to the underlying hardware resources by designing dynamically reconfigurable accelerators in the cloud. Two cloud service: Reconfigurable IPs as a Service (RIPaaS) and Reconfigurable Region as a Service (RRaaS) are considered in this paper. Similarly, the connection and communication between the accelerators and the reconfigurable control will not be efficient without the use of Network-on-Chip (NoC). In order to address these issues, we propose a NoC based virtualized accelerators for cloud computing. Reconfigurable accelerators communicate their status and exchange data through the routers connected to them. A Hypervisor is used as an intermediate level between the physical reconfigurable hardware and application layer to manage the FPGA resources. A 2x2-mesh NoC based reconfigurable accelerators for image analysis and matrix computation are implemented and tested showing a promising result for more scalable systems in cloud computing.