With the blossom of virtualization and cloud computing, virtualized systems can be found from small companies to service providers and big data centers. All of them use this technology because of the many benefits it has to offer, such as a greener ICT, cost reduction, improved profitability, uptime, flexibility in management, maintenance, disaster recovery, provisioning and more. The main reason for all of these benefits is server consolidation which can be improved with dynamic resource allocation techniques. Out of the resources to be allocated, memory is one of the hardest and it needs proper planning and ideally good predictions and proactivity. Many attempts have been made to approach this problem, but most of them are using traditional statistical mathematical methods. In this thesis the application of discrete Bayesian networks is evaluated, to offer probabilistic predictions on system utilisation with focus in memory. Bayllocator is built to provide proactive dynamic memory allocation based on the Bayesian predictions, for a set of virtual machines running in a single hypervisor. The results show that with proper tuning Bayesian networks are capable of providing good predictions for system load, and increase performance and consolidation of a single hypervisor. Their modularity gives great freedom for experimentation, and with small changes in nodes and the causal relationship of the nodes, different results can be obtained.