Prediction of virtual networks substrata failures

In a Virtual Network Environment (VNE), a failure in the substrate network will affect the many virtual networks hosted by the substrate network. To minimize un-predicted failures, maximize system performance, efficiently use resources and determine how often failures may occur, we must be able to predict failure occurrence. In this paper, we present a prediction mechanism to forecast the Time-To-Failure (TTF) of the VNE components based on time series data. In addition, we use supervised learning based on a Support Victor Regression (SVR) model to predict future failures in the VNE. The prediction can be used to establish a tolerable maintenance plan in the event of substrate and virtual network failure. Failure prediction can be used to enhance virtual network (VN) dependability by forecasting the failure occurrences in the substrate network using runtime execution states of the system and the history of observed failures.