A company offering a cloud-computing service provides any customers request of computing resources (processors, virtual memory, disk space). The company has to satisfy two competitive needs: one is to provide the proper hardware available to fulfil any requests and the necessity to not leave a large number of unused machines.
In order to achieve the requirements, it is crucial to have a rational prediction exploiting all the information available and indicating the likelihood of receiving a set of requests in a specific time period. It is possible to model the flux of customers requests and infer from archived data the rate of arrival of a request, and its features.
The Bayesian approach is an extremely powerful tool to model the prediction and provides results of clear interpretation. By studying the history of all requests and knowing the resources available, Pangea Formazione has developed a model that simulates (using Monte Carlo techniques) scenarios for the near future requests. In addition, the tool has a graphical interface allowing the user to estimate the time interval where the currently available resources will be fulfilled and the likelihood of meeting the requests in a given time.
The company supported by this predictive tool can rationally determine when it is time to enhance the machine fleet. Moreover, the tool allows to quantitatively manage the risk of delaying the service delivery to specific customers and compare it to the cost of a purchase of other machines.