A challenging issue in the development of parallel implementations of Data Stream Processing patters is to be able to reduce the costs of dynamic memory allocation. In this work we study two parallel patterns for sliding-window computations and we discuss different implementation variants related to how dynamic memory is managed. Slides available here