Accurate Prediction of Workflow using Dual-Stage Learning to Reduce Task Execution Time

Abstract

As the number of cloud data centres continues to expand rapidly, one of the biggest worries is how to keep up with the energy demands of all these new servers without negatively impacting system dependability and availability or raising the price of power for service providers. Workflow task performance prediction for variable input data is crucial to several methods, including scheduling and resource provisioning. However, it is challenging to create such estimations in the cloud. The suggested system's two-stage forecasts and parameters that account for runtime data, allow for very precise predictions. The workflow is smooth, and obviously the task execution time is adequate. This strategy beats the state-of-the-art prediction techniques, as shown by empirical data. It is demonstrated that the models of this form, predicting workflow for a given cloud, can be easily transferred to other clouds with little effort and error.