Punctual Cloud: Unbinding Real-time Applications from Cloud-induced Delays

Cloud computing has become a prominent technology for the computing paradigm in various industrial sectors nowadays. For most industrial applications to perform in real-time, the support of periodic computing is required. However, it remains a challenge when the computing is executed in a cloud, since both the network connection and the cloud environment are uncertain. In this paper, we propose a new architecture to deploy real-time applications in the cloud. We call it punctual cloud. We detail the implementation and demonstrate how punctual cloud is deployed in a cloud-native manner on Kubernetes. We evaluate the system's performance with a real-time resource allocation problem and show that, compared to a system without punctual cloud, which has maximum 40% punctual deliveries, our proposed architecture can attain over 90% responses to be delivered punctually for the application, while also being capable of remedying the performance degradation caused by long and uncertain response delays in the system.