USING CLOUD COMPUTING AS A HPC PLATFORM FOR EMBEDDED SYSTEMS

Abstract: The aim of this paper is to propose a first prototype of a High Performance Computing platform for robotics, more specifically in the area of elder people care and rehabilitation. Apart from getting better performance with this platform, one main objective is to save as much as possible the robot’s power consumption. The system proposed is a Cloud-based platform, which main advantages are those of infinite resources, dynamic scalability and better resource utilisation. The different elements and technologies involved in a Cloud will be presented, as well as explaining the whole building process. The Cloud was implemented using Openstack middleware and KVM with hardware-based virtualisation. Our platform was validated for HPC with some example algorithms such as a Laplace Equation solver, and a Object detector. For the latter one, we also compared its performance against example embedded system hardware, simulating a real case. The results obtained show that our Cloud can perform better than the embedded system despite the networking overheads.