Many scientific devices and equipment generate streaming data at a very fast pace. The ability to transform, clean, filter, summarize, analyze and visualize such fast data streams in an on-line and real-time fashion is desirable in many applications. In this talk, we will review some recent research in fast data stream processing, with the focus on the supporting technologies for data stream processing systems, such as query and programming models, memory management, job parallelization, fault tolerance and so on. We will also look into several existing stream processing systems and analyze their pros and cons.