Peer-to-peer systems have gained momentum over the last years. This work was motivated by the necessity to analyze such systems. It proposes a distributed monitoring system, P2PMonitor, that has alerters specialized in local monitoring of entities. These alerters encode the basic events detected into streams of (Active)XML documents. A method for filtering efficiently such streams has been introduced. Also, an algorithm has been proposed for the detection of parts of a new monitoring task that are already supported by the system. Complex and efficient stream processors are needed at the heart of P2PMonitor. The work shows how to build them from views over active documents and proposes a maintenance algorithm that scales and reduces the computation time. Modeling the applications is important for being able to monitor them. At their origin, the business processes have been mainly operation-centric, but recently, business artifacts have been proposed and seem well adapted for the specification of data-centric applications. The proposed artifact model is based on active documents and captures: their state, evolution, interactions and history.