Autonomic computing architecture for real-time medical application running on virtual private cloud infrastructures
Yong woon Ahn, Albert M. K. Cheng · ACM SIGBED Review · 2013
Cloud computing with virtualization technologies has become a huge trend which attracts academia and information technology industries because of its cost-efficiency. It has changed paradigms of development, release, and maintenance of diverse types of software and service. However, this big movement has not been applied to real-time applications yet because deploying real-time applications on the cloud infrastructures arouses many controversies because of numerous uncertainties from sharing physical resources. It is not trivial to apply conventional scheduling techniques for real-time systems to cloud infrastructures without further considerations. All virtual machines (VMs) must be controlled by the Virtual Machine Monitor (VMM) which is centralized and has a primary role to share physical resources fairly with all VMs in the same physical machine (PM). For best-effort applications, this fair resource sharing policy works well and end-to-end Quality of Service (QoS) is promised by the service level agreement (SLA) with reasonable delay windows. However, for real-time applications with aperiodic hard- and soft-real-time tasks, this mechanism has serious weaknesses. Although VMMs of most cloud infrastructures have auto-scaling and load-balancing mechanisms, these are unpredictably slow to accept urgent aperiodic-real-time tasks because of its fair resource sharing policy. Therefore, it is necessary to propose another approach to satisfy deadline constraints of real-time tasks transferred from remote locations. In this research, we focus on cloud medical applications processing sensitive patient data with deadline constraints. We propose feasible solutions to reserve computing and networking resources with an autonomic computing architecture in the virtual private cloud infrastructures.