Management activities are based on the state of distributed system components, relations of these components, and their behaviour. Since management policies are applied across an abstraction of distributed systems, the quality of decisions is dependent on the representation fidelity of the system real state. Obviously, the data collection process updating the abstract representation of real distributed systems has a major significance to carry out relevant management decisions. Existing approaches promote a fixed polling frequency for all components. We present shortcomings of these issues and propose adapted polling formulae considering the behavioural parameters of managed components. Models for the flexible polling frequency and for the frequency adaptor are presented. Based on these models, we propose adapted formulae considering both the availability features and dynamics features in a linear and exponential approach. Policies for applying these formulae are proposed with respect to the component grouping and frequency classification. Implementation aspects are discussed.