Resilience Assessment of Cloud Video Transcoding Services

In the digital era, cloud-based video transcoding services are essential for platforms like Netflix and YouTube, relying on virtual machines for efficient processing. While prior research has primarily explored performance evaluation, limited attention has been given to the resilience of these services against emerging disruptions. This paper aims to predict the processing rate of video transcoding services to evaluate the resilience of small, medium, and large VMs within a cloud environment by applying four regression models. Findings reveal that a mixed regression model incorporating covariates, their interactions, and polynomial terms provides the most accurate prediction of performance and resilience. Among the VMs, the small VM exhibits the highest resilience, sustaining robust performance across varying conditions, with mean time to failure identified as the key factor in its sustained operation. These insights underscore the significance of predictive models for forecasting resilience, supporting proactive decisions for preventive maintenance and fault-tolerant design to reduce cloud service downtime.