Virtual FPGA Placement with an Efficient Ant Colony Optimization

Abstract

Virtualization allows integrating Field Programmable Gate Arrays (FPGAs) into a resource pool at the infra-structure layer. So as to improve the FPGA resource utilization while ensuring the quality of service, a virtual FPGA (vFPGA) Scheduling algorithm has been presented in our early work. At the meantime, we noticed that the initial deployment of vFPGAs has obvious effect on resource utilization ratio. Finding an optimal deployment of vFPGAs onto FPGAs which can be summed up in virtual FPGA placement (VFP) problem is a NP-hard problem. With a widespread of reconfigurable cryptographic resource pool, regarded it as a combinatorial optimization problem have offered higher efficiency than linear programming (LP) problem. In this paper, an optimized ant colony optimization (ACO) algorithm, where given ants the ability to perceive resource status, is presented to achieve the VFP goal. Finally, CloudSim toolkit is extended to evaluate our solution through simulations on synthetic workloads. The obtained results show that our algorithm can reduce the number of active FPGAs by improving the resource utilization.