Semantic resource allocation with historical data based predictions

One of the most important issues for Service Providers in Cloud Computing is delivering a good quality of service. This is achieved by means of the adaptation to a changing environment where different failures can occur during the execution of different services and tasks. Some of these failures can be predicted taking into account the information obtained from previous executions. The results of these predictions will help the schedulers to improve the allocation of resources to the different tasks. In this paper, we present a framework which uses semantically enhanced historical data for predicting the behavior of tasks and resources in the system, and allocating the resources according to these predictions.