Automatic Resource Scaling for Medical Cyber-Physical Systems Running in Private Cloud Computing Architecture

Abstract

Cloud computing and its related virtualization technologies have become one of dominant trends to deploy software, compute difficult problems, store different types of data, and stream real-time video and audio. Due to its benefits from cost-efficiency and scalability to maintain server solutions, many organizations are migrating their server applications running on physical servers to virtual servers in cloud computing infrastructures. Moreover, cloud computing has enabled mobile and battery-powered devices to operate without strong processing power and large storage capacity. However, it is not trivial to use this trendy technology for medical Cyber Physical Systems (CPSs) which require processing tasks’ requests to send instructions to the local actuator within specified deadlines. Since a medical CPS device monitoring a patient’s vital signs may not have a second chance to recover from an erroneous state, achieving cost-efficiency with higher resource utilization in cloud computing may not be the ultimate goal to configure the healthcare IT infrastructure with medical CPS devices. In this paper, we focus on private cloud infrastructures with the fair resource sharing mechanism in order to run medical CPS applications. First, we introduce our medical CPS device model used for designing our cloud infrastructure following the Integrated Clinical Environment (ICE) standard developed by the Medical Device Plug-and-Play (MDPnP) project. Second, we investigate limitations to deploy CPS applications using existing auto-scaling mechanisms. Finally, we propose our novel middleware with a virtual resource sharing mechanism inspired by autonomic computing, and present its performance evaluation results simulated in the OpenStack private cloud.