The four LHC experiments rely on experiment specific services running on machines mainly located at CERN. Some of these services have been rated by the experiments as very critical: any loss or degradation of performance has a major impact on the experiment's production and analysis activities. It is therefore important to provide a reliable and robust operational environment. In this work we describe the strategy based on service deployment, security and virtualization adopted to enhance the reliability of ATLAS and CMS central services.