A Probabilistic Learning Approach for Predicting Application Launches in Cloud Computing Architectures

Desktop and application virtualization suffers from delay. Opening sessions and remote applications implies to load a non-negligible amount of data, parameters and services. This can be solved by predicting the future activities of the users and pre-loading the required resources. We thus propose a complete approach that automatically discovers the periodical patterns of the users and that automatically builds a probabilistic model of user's behaviour. A kernel density estimator is exploited to estimate the probability density function of an application to be launched by a user. Using the probabilistic approach allows to predict the application a user will open and to reduce the launching time. The efficiency of the proposed approach has been verified by an implementation in a virtualization tool under real operating conditions.