The Internet of Things (IoT) is shaping to a worldwide network of networks consisting of billions of interconnected heterogeneous sensor/actuator-equipped devices (denoted as 'things' or 'smart objects'), which are expected to exceed 50 billions by 2020. Smart objects, which will be pervasively deployed, are constrained devices with (i) limited processing power and available memory and (ii) limited communication capabilities, in terms of transmission rate and reliability. Future Smart-X applications, such as Smart Cities and Home Automation, will be fostered by the use of standard and interoperable IP-based communication protocols that smart objects are going to implement, by simplifying their development, integration, and deployment. Smart-X applications will significantly differ from traditional Internet services, in terms of: (i) the number of data sources; (ii) rate of information exchange; and, (iii) need for real-time processing. Because of these requirements, such services are denoted as 'Big Stream' applications, in order to distinguish them from traditional Big Data applications. In this paper, we present an implementation of a novel Cloud architecture for Big Stream applications based on standard protocols and open-source components, which provides a scalable and efficient processing platform for IoT applications, designed to be open and extensible and to guarantee minimal latency between data generation and consumption. We also provide a performance evaluation based on experimentation in a real-world Smart Parking scenario, to assess the feasibility and scalability of the proposed architecture.