|To date, there is limited literature available on the consideration of system failures in exploring the performance of integrated architecture in Medical Internet of Things (MIoT) devices, fog computing, and cloud computing. MIoT devices refer to internet-connected devices that are designed to be portable and equipped with sensors, processors, and wireless communication
capabilities that enable them to collect and transmit health data over the internet. The limited literature available on the consideration of system failures in exploring the integrated architecture of MIoT devices, fog computing, and cloud computing is problematic because the failure of devices and infrastructure may result in negative outcomes, including
the potential death of patients or users.
To overcome the above challenge, this study developed analytical models of a robust Internet
of Things (IoT) enabled healthcare monitoring system that incorporates system failures. The analytical models were developed using queueing theory which is a mathematical theory that deals with the study of waiting lines or queues and provides a framework for analyzing the performance of systems. The robust Internet of Things (IoT) enabled healthcare monitoring system consists of computing resources hosted on a platform that consists of fog nodes, private and public clouds having virtual machines that process requests.
To ensure robustness of the monitoring system, the developed analytical model takes into consideration system/component failures. The monitoring system also has the ability to recover from unexpected failures and maintain performance in the face of adverse events. The study investigated the effect of varying the arrival rate and system load on mean response time for healthcare data requests when failure is incorporated and when failure is not accounted for. The numerical results obtained show that the mean response time of healthcare data requests is the same when the arrival rate or load is low for situations when failure is incorporated in the system model and when it is not, however when the arrival rate or load is high, the mean response time is higher when system failure is incorporated in the system model.
It is observed that increase in the number of virtual machines from 3 to 4 generally leads to a decrease in mean response time of about 14ms when the arrival rate of packets in the system is about 9000 requests/second and the system is 90% utilized. It is noted that when the probability that packets are forwarded to the fog layer is increased from 0.6 to 0.7, and the probability that packets are forwarded to the private cloud reduced from 0.4 to 0.3, the mean response time is reduced. Since system failure has significant impact on system performance especially at high arrival rate values and load, it should be incorporated in developing analytical models for mean response for cloud-IoT architectures that are used for healthcare monitoring.